Journal April 2011 Edition

International journal of scientific
and engineering research
IJSER
Volume 2, Issue 4, April 2011
ISSN 2229-5518
Research
Publication
IJSER
http://www.ijser.org
http://www.ijser.org/xplore.html
http://www.ijser.org/forum
E-mail: [email protected]
International Journal of Scientific and Engineering Research (IJSER)
Journal Information
SUBSCRIPTIONS
The International Journal of Scientific and Engineering Research (Online at www.ijser.org) is published
monthly by IJSER Publishing, Inc., France/USA/India
Subscription rates:
Print: $50 per issue.

To subscribe, please contact Journals Subscriptions Department, E-mail: [email protected]
SERVICES
Advertisements
Advertisement Sales Department, E-mail: [email protected]
Reprints (minimum quantity 100 copies)

Reprints Co-ordinator, IJSER Publishing.
COPYRIGHT
Copyright2011 IJSER Publishing, Inc.
All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in
any form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise, except as de-
scribed below, without the permission in writing of the Publisher.
Copying of articles is not permitted except for personal and internal use, to the extent permitted by national copy-
right law, or under the terms of a license issued by the national Reproduction Rights Organization.
Requests for permission for other kinds of copying, such as copying for general distribution, for advertising or pro-
motional purposes, for creating new collective works or for resale, and other enquiries should be addressed to the
Publisher.
Statements and opinions expressed in the articles and communications are those of the individual contributors and
not the statements and opinion of IJSER Publishing, Inc. We assumes no responsibility or liability for any damage
or injury to persons or property arising out of the use of any materials, instructions, methods or ideas contained
herein. We expressly disclaim any implied warranties of merchantability or fitness for a particular purpose. If ex-
pert assistance is required, the services of a competent professional person should be sought.
PRODUCTION INFORMATION
For manuscripts that have been accepted for publication, please contact:
Contents
1. Haptic Training Method Using a Nonlinear Joint Control

Jarillo-Silva Alejandro, Domnguez-Ramrez Omar A., Parra-Vega Vicente.1-8
2. Prevalence of Eating Disorder among Female Students of Tonekaban University

Ida Fadavi, Roxanna Ast.5-13
3. Impact Of Leverage On Firms Investment Decision

Franklin John. S, Muthusamy .....13-28
4. Effect of Nanofluid Concentration on the Performance of Circular Heat Pipe

M. G. Mousa ..29-36
5. Enhancement of person identification using Iris pattern

Vanaja roselin.E.Chirchi, Dr.L.M.Waghmare, E.R.Chirchi....37-42
6. Mn DOPED SnO2 Semiconducting Magnetic Thin Films Prepared by Spray

Pyrolysis Method
K.Vadivel, V.Arivazhagan, S.Rajesh ..43-47
7. A novel comprehensive method for real time Video Motion Detection

Surveillance
Sumita Mishra, Prabhat Mishra, Naresh K Chaudhary, Pallavi Asthana.....48-51
8. Handoff Analysis for UMTS Environment

Pankaj Rakheja, Dilpreet Kaur, Amanpreet Kaur.....52-55
9. Performance and Emission Characteristics of Stationary CI Engine with Cardnol

Bio fuel blends
Mallikappa, Rana Pratap Reddy, Ch.S.N.Muthy .....56-61
10. Morphological Space and Transform System

Ramkumar P.B, Pramod K.V...62-66
11. Bus Proximity Indicator - (An Intelligent Bus Stop)

Prof. A.P. Thakare, Mr. Vinod H. Yadav ...67-70
12. Designing aspects of Artificial Neural Network controller

Navita Sajwan, Kumar Rajesh ....71-75
Contents
13. Handwritten Character Recognition using Neural Network

Chirag I Patel, Ripal Patel, Palak Patel ..76-81
14. Improved Performance of M-ary PPM in Different Free-Space Optical Channels

due to Reed Solomon Code Using APD
Nazmi A. Mohammed, Mohammed R. Abaza and Moustafa H. Aly .82-85
15. Electricity Sector Restructuring Experience of different Countries

Archana Singh, Prof. D.S.Chauhan ...86-93
16. Significant Role of Search Engine in Higher Education

Rahul J. Jadhav, Dr. Om Prakash Gupta, Usharani T. Pawar ...94-98
17. Feature Selection for Cancer Classification: A Signal-to-noise Ratio Approach

Debahuti Mishra, Barnali Sahu..99-105
18. Towards Analysis of Vulnerabilities and Measuring Security Level of Copmputer

Network Life Cycle
Irshad Ahmad Mir, Mehraj-U-Din Dar, S.M.K Quadri106-111
19. Text Independent Speaker Identification In a Distant Talking Multi-Microphone

Environment Using Generalized Gaussian Mixture Model
P. Soundarya Mala, Dr. V. Sailaja, Shuaib Akram112-116
20. Novel Object Removal in Video using Patch Sparsity

B. Vidhya, S. Valarmathy..117-121
21. New formula of Nuclear Force

Md. Kamal uddin...122-128
22. Impact Fatigue Behaviour of Fully Dense Alumina Ceramics with Different
Grain Sizes
Manoj Kumar Barai, Jagabandhu Shit, Abhijit Chanda, Manoj Kr Mitra.129-133
23. Distributed Generation Planning Optimization Using Multiobjective

Evolutionary Algorithms
Mahmood Sheidaee, Mohsen Kalantar......134-139
Contents
24. Marketing Of Asian Countries as Tourist Destination- Comparative Study of

India and Malaysia
Dr Reshma Nasreen, Nguyen Toan Thang140-156
25. Clinical and Computational Study of Geometry & Heamodynamics of Arterial

Stenosis
Krittika Dasgupta, Abhirup Roy Choudhury, Abhijit Chanda, Debabrata Nag.157-162
26. Segmentation Techniques for Iris Recognition System

Surjeet Singh, Kulbir Singh..163-170
27. Investigation of X-ray Plasmon Satellites in Rare Earth Compounds

Ajay Vikram Singh, Dr. Sameer Sinha171-175
28. Improvement of Power System Stability by Simultaneous AC-DC Power

Transmission
T.Vijay Muni, T.Vinoditha, D.Kumar Swamy...176-181
29. A Comprehensive Distributed Generation Planning Optimization with Load

Models
Mahmood Sheidaee, Mohsen Kalanter...182-187
30. Positive Impact of Mobility Speed on Performance of AODV at Increased

Network Load
Rajesh Deshmukh, Asha Ambhaikar..188-192
31. Development and Physical, Chemical and Mechanical Characterization of

Doped Hydroxyapatite
Promita Bhattacharjee, Howa Begam, Abhijit Chanda193-200
32. A chart for general and complicated data visualization

M Siluvairajah201-202
33. Application of Reliability Analysis: A Technical Survey

Dr. Anju Khandelwal203-210
34. An Innovative Quality Of Service (Qos) Based Service Selection For Service
Orchestration In Soa
K.Vivekanandan, S.Neelavathi211-218
Contents
35. Design of Boost Circuit for Wind Generator

N.Prasanna Raj, M.Mohanraj, Rani Thottungal219-224
36. A Novel Method for Fingerprint Core Point Detection

Navrit Kaur Johal, Amit Kamra...225-230
37. Robust Impulse Eliminating Internal Model Control of Singular Systems

M.M. Share Pasand, H.D. Taghirad231-238
38. An Integrated Partial Backlogging Inventory Model having weibull demand

and variable deterioration rate with the effect of trade credit
P.K. Tripathy, S. Pradhan.239-243
39. Survey of Current and Future Trends in Security in Wireless Networks

Vikas Solomon Abel..244-249
40. Anomalous hydrogen production during photolysis of NaHCO3 mixed water

Muhammad Shahid, Noraih Bidin, Yacoob Mat Daud, Muhammad Talha250-254
41. Peripheral Interface Controller based the Display Unit of Remote Display
System
May Thwe Oo.255-263
42. Training and Analysis of a Neural Network Model Algorithm

Prof Gouri Patil..264-268
43. Wireless Sensor Network: A Review on Data Aggregation

Kiran Maraiya, Kamal Kant, Nitin Gupta..269-274
44. Slant Transformation As A Tool For Pre-processing In Image Processing

Nagaraj B Patil, V M Viswanatha, Dr. Sanjay Pande MB....275-281
45. Separation of concerns in VoiceXML applications

Sukhada P. Bhingarkar.282-285
46. Error of Approximation in Case of Definite Integrals

Rajesh Kumar Sinha, Satya Narayan Mahto, Dhananjay Sharan...286-289
Contents
47. An Assessment model for Intelligence Competencies of Accounting

Information Systems
Mehdi Ghazanfari, Mostafa Jafari, Saeed Rouhani...290-296
48. Processing of Images based on Segmentation Models for Extracting

Textured component
V M Viswanatha, Nagaraj B Patil, Dr. Sanjay Pande MB297-303
49. An Adaptive and Efficient XML Parser Tool for Domain Specific Languages
W. Jai Singh, S. Nithya Bala.304-307
1
International Journal of Scientific & Engineering Research Volume 2, Issue 4, April-2011
ISSN 2229-5518
Haptic Training Method Using a Nonlinear

Joint Control
Jarillo-Silva Alejandro, Domnguez-Ramrez Omar A., Parra-Vega Vicente
Abstract There are many research works on robotic devices to assist in movement training following neurologic injuries such
as stroke with effects on upper limbs. Conventional neurorehabilitation appears to have little impact on spontaneous biological
recovery; to this end robotic neurorehabilitation has the potential for a greater impact. Clinical evidence regarding the relative
effectiveness of different types of robotic therapy controllers is limited, but there is initial evidence that some control strategies
are more effective than others. This paper consider the contribution on a haptic training method based on kinesthetic guidance
scheme with a non linear control law (proxybased second order sliding mode control) with the human in the loop, and with
purpose to guide a human users movement to move a tool (pen in this case) along a predetermined smooth trajectory with
finite time tracking, the task is a real maze. The path planning can compensate for the inertial dynamics of changes in direction,
minimizing the consumed energy and increasing the manipulability of the haptic device with the human in the loop. The
Phantom haptic device is used as experimental platform, and the experimental results demonstrate.
Index TermsDiagnosis and rehabilitation, haptic guidance, sliding mode control, path planning, haptic interface, passivity and
control design.
1 INTRODUCTION
I N the least decade the number of patients who have

suffere accidents stroke or traumatic brain injuries have
increase considerably [1]. The central nervous system
ing robotic instrumentation is a tool that accomplishes the
goal of the patients recovery.
damage can lea to impaired movement control upper 1.1 Justification

extremities, which ar facing major difficulties in relation As care is decentralized and moves away the inpatient
to the activities of daily life. Several studies showed that settings to homes, the availability of technologies can
the rehabilitation therapy which is based on motion- provide effective treatment outside, the acute care of the
oriented tasks repetitive, it help to improve the move- hospital would be critical to achieve sustainable man-
ment disorder of these patients [2], [3]. Unfortunately agement of such diseases. In the field of neuromotor re-
repeatability therapy requires consistency, time for physi- habilitation, skilled clinical managers and therapists can
cians and therefore money. achieve remarkable results without using technological
Conventional neurorehabilitation appears to have little tools or with rudimentary equipment, but such precious
impact on impairment on spontaneous biological recov- human capital is in very short supply and, in an case, is
ery. Robotic neurorehabilitation has potential for a greater totally insufficient to sustain the current demographic.
impact on impairment due to an easy deployment, its Robots are particularly suitable for both rigorous test-
applicability across of a wide range of motor impairment, ing and application of motor learning principles to neuro-
its high measurement reliability, the capacity to deliver rehabilitation. Computational motor control and learning
high dosage of training protocols and high intensity in the principles derived from studies in healthy subjects are
exercises. This situation economic rehabilitation. With the introduced in the context of robotic neurorehabilitation.
purpose of enhance the relationship between outcome There is an increasing interest in using robotic devices to
and the costs of rehabilitation robotic devices are that provide rehabilitation therapy following neurologic inju-
being introduced in clinical rehabilitation [4], [5]. Rehabil- ries such as stroke and spinal cord injury. The general
itation using robotic devices not only has had an impor- paradigm to be explored is to use a robotic device to
tant contribution in this area but also has introduced physically interact with the participants limbs during
greater accuracy and repeatability of rehabilitation exer- movement training, although there is also work that uses
cises. Accurate measurement quantitative parameter us- robot that do not physically contact the patients [6].
1.2 The Problem

Jarillo-Silva Alejandro, University of the Sierra Sur, Mxico, Oaxaca, E- Some of the problems presented by a motor rehabilitation
mail: [email protected]
Domnguez-Ramrez Omar A. University of Hidalgo State, Mxico, E- of patients who have suffered a brain injury, is the time
mail: [email protected] rehabilitation takes place, the cost that this entails, poor
Parra-Vega Vicente, Robotics and Advanced, Manufacturing Divi- information that the specialist has to determine the diagno-
sion, CINVESTAV Saltillo, Mxico, E-mail: vpar-
sis of a patient in rehabilitation, and the lack of platforms
[email protected]
for rehabilitating patients who cannot assist to the hospital
IJSER 2011
2
ISSN 2229-5518
and require continued rehabilitation.
1.3 Our Proposal 3.2 Haptic Device Dynamics

One technique used to solve the problem of generating Consider a mechanism of articulate links (PHANToM 1.0
rehabilitation platforms using robotic systems is haptic [11]), with n revolute joints described in the generalized
guidance, this technique is based on the use of haptic de- join coordinates ( q T , q T ) T R 2 n . Physically, the robot is
vices, which allow human-machine interaction. These hap- never in touch with a physical object, thus robot dynamics
tic devices are programmed under certain considerations, in free motion describe properly the haptic device, as fol
such as considering the safety of the patient during the lows
interaction, the physiology of the patient in order to gener-
ate tracking trajectories that allow proper rehabilitation, D ( q) q (C ( q, q ) B ) q G (q ) (1)
and so on. In this paper is proposed to generate a trajectory 3x3
based on the solution of mazes in 2D and the haptic device Where D ( q ) R denotes a symmetric positive de-
guides the patient under a control law, which is designed finite inertial matrix, C ( q , q ) R 3 x 3 is a Coriolis and centri-
with certain characteristics that allow coupling among the petal forces matrix, G ( q ) R 3 x1 models the gravity forces
patient, the device and the trajectory. The haptic device is vector, B R 3 x 3 denotes a symmetric positive definite
equipped with optical encoders, this allows obtaining data viscous coefficient matrix, and R 3 x1 stands for the
such as position, velocity, acceleration, force and torque, torque input. In our case the human operator is driving the
these data in combination with other data allow to the system, therefore c h , and c R 3 x1 stands for
specialist to generate a clinical diagnosis, which has thera- the guidance control, and h J T f h denotes the human-
peutic support based on the patients movements in eac haptic interaction, where J T stands for the transpose
exercise performed on the platform. Jacobian of the haptic device, an fh represents the human
performance force vector.
1.3 Organization 3.3 Guidance Control Law
In section 2, we introduce the human-haptic interaction, The design of a control strategy applied to haptic guided
including the haptic scheme, the dynamic model of the must meet certain objectives such as: giving stability in
haptic device and the guidance control law implemented in human-machine interaction, presenting high robustness
the experimental platform. The description task and the against disturbances generated by the human operator and
path planning for guiding to the experiments are given in currently meeting the objective to rehabilitate, this can be
section 3; experimental results are discussed in 4. Finally, corroborated with an analysis clinical examined by special-
we presen the conclusions in section 5. ists.
3.4 Properties of Euler-Lagrange Systems
3 HUMAN-HAPTIC INTERACTION
The dynamic equation for manipulator robots (1) has in-
The haptic feedback has also been developed for task strusting follow properties [12]:
oriented biofeedback studies [7]. Haptic interfaces allow
the patient to interact with and to manipulate a virtual There exists some positive constant such that
object or real. Results [8] have shown that haptic informa-
tion provides knowledge of results (KR) and kinaesthetic D (q) I q R n
fed back, sensation that are important for task perfor-
mance. Where I denotes the nxn identity matrix. The
D (q )1 exist and this is positive definite.
3.1 Haptic Guidance Sheme
The matrix c( q , q ) have a relationship with the
For the human participation in the system, there are two
inertial matrix as:
haptic guidance applications: active and passive. The first
1
allows the user to guide the robotic system (haptic device) q T D ( q ) C ( q , q ) q 0 q , q R n
in the workspace to explore. However, in the passive hap- 2
tic guidance, the user is guided in the workspace with the ((
The Euler-Lagrange system have the total energy
position and velocity control strategies. On the other hand,
as:
kinesthesis is the human sense of position and movement,
which is created from proprioceptive cues arising from
k u (2)
Where k is the kinetic sum for each coordianate, and
receptors in the joints and muscles [9]. Thanks to kinesthet-
ic memory, or the ability to remember limb position and
u are the potencial energy respectively.
The kinetic energy is obtained by:
velocity, humans have a remarkable ability to remember
positions of their limb quite accurately and for long periods 1 n 1
[10]. k mi vi2 q T D(q)q (3)
2 i1 2
IJSER 2011
3
ISSN 2229-5518
Where mi R are the i mass of the i link, vi R n are With the Lemma 1, and the dynamic properties, we can
the i velocity of the i link. follow a control law design. In this paper we obtain the
The potencial energy is obtained by: control technique via Lyapunov theory, dynamical and
passivity properties.
n
u mi hi g (4) 4 DESIGN OF A NONOLINEAR JOINT
i 1
CONTROL BASED ON PASSIVIT
n
Where hi R are the i height of the i link respect to Given a designated trajectory qd (t ) , can be considered as a
mass center and g is a gravitational constant. By differen- Lyapunov function as follows [14]:
tiating (2) we obtain 1 T
V ( s) s D(q ) s s T K L tanh( s )dsds c (10)
1 2
q D (q) q q T D (q )q q T G ( q)
T
2 (5) Where c >>1, s s1, s 2,.., sn T denote dynamic error,

the function tanh( s ) tanh( s1),.., tanh( sn ) and K L are a
T
1
q T C (q, q ) G ( q) q T D (q) q q T G (q ) symeyric positive definite matrices.
2
Given that (10) is a positive semidefinite function, ob-
q
T
tain the derivative of the Lyapunov candidate function, so
that is follows [1]:
From the passivity property we have that:
1 1 1
t V ( s ) s T D (q ) s s T D ( q ) s sT D ( q ) s (11)
T (6) 2 2 2
V ( x) V (0) y ( s )u ( s )ds T
0
s K L tanh( s ) ds
Where V (x) is a storage function, y(s) is the output, Simplifyng (11)
and u(s) is the input of the system, and s is a variable
1
change. Using (5) for the Euler-Lagrange system, energy V ( s ) s T D (q ) s s T D (q ) s s T K L tanh( s )ds (12)
function as the storage function, and we have the passiv- 2
ity property as:
As saw in (12), we applied lemma 1, and use the follow
t
result:
(t ) (0) q T dt (7)
0 D (q) s C (q, q ) K d s (13)
Where q is the output, and is the input of the sys
tem. Where are the system dynamics
3.5 Passivity of the Dynamic Error
D(q )qr C (q, q )q r G (q ) F (q ) (14)
The described passivity property presents an interesting
property in the regulation problem on Euler-Lagrange Yr (q, q.q r , qr )
systems, however this property can be extended for the
trajectory tracking problem solution, [13] drive this passive Where q r and q are the nominal references,
erro dynamic property, as follow for: nxp r
Yr (q, q.q r , qr ) R is the regressor of known nonli-
p
near terms R is the vector of unknown parameters,
D(q ) s C (q, q ) K d ( q, q )s 0 (8) so that (13) is as follows:
Where s denotes an error signal that we want to drive

D (q )s Yr (q, q.q r , qr ) C (q, q ) K d s (15)
to zero, k d ( q , q ) k d ( q , q )T 0 is a damping injection ma-
trix. Based on this property and the skew-symmetric prop- Using (15) and (12) as follows:
erly we follow the next lemma:
Lemma 1: The differential equation: V ( s) sT Yr (q, q.qr , qr ) C (q, q ) K d s (16)
1
D(q ) s C (q, q ) K d (q, q )s (9) sT D (q ) s sT K L tanh( s )ds
2
Where D(q) and K d are positive matrices and V ( s ) s T Yr ( q, q.q r , qr ) K d s s T C (q, q ) s
C (q, q )
satisfies (8) it defines an output strictly passive
operator
.

d : s . Consequently, if 0 we have
1 T
s D ( q ) s sT K L tanh( s )ds (17)
sL 2
2
IJSER 2011
4
ISSN 2229-5518
Applying the antisymmetric property in (17), it reduces

to the following expression:

V (s) sT Yr (q, q.qr , qr ) Kd s KL tanh(s)ds (18)
For asymptotic convergence (Lyapunov stability theo-

rem) is requiredV ( s ) 0 , in this case a function is de- Fig. 1. Haptic guidance system[15]
signed V as follows:
performed with the joint space required to use models
V ( s ) s T K D s 0 (19) IKMP (Inverse Kinematic Model Position) and IKMV
(Inverse Kinematic Model Velocity), so qr is obtained
Yr ( q, q.q r , qr ) K d s which represents the vector of desired joint positions and
V ( s ) s K D s s
T T
(20) qr is the vector of desired joint velocity, in this way is
K L tanh( s )ds
obtained qp which represents the vector of real joints
In (20), can be reduced to the following expression: position and q r represents the vector of real joints
velocities. By getting the joint position errors and velocity
Yr ( q ,
q .q r ,
q r ) K s
d
articulated errors , they are used in the joint control to
sT K D s sT (21) generate the vector of torques. The patient is manipulated
K L tanh( s )ds
by the haptic device, but the patient disrupts desired
tracking trajectory, the perturbation (in position and
Solving for in (21) is obtained: velocity) is obtained by reading optical encoders mapping
is performed after the joint space to the operational space
Yr ( q, q.q r , qr ) K d s K D s to display the operational positions and velocities.
(22)
L
K tanh( s ) ds 5.1 Phantom Haptic Device Platform
Where K D kd , so that K d s K D s can be written as The experimental platform consists of a PC running at 1.4
k ds s , therefore the control law for systems n degrees of GHz with two AMD Athlon processor, 1 Gb RAM, 64 Mb
freedom can be written as: NVIDIA GeForce 4Mx and Visual C++ as a programming
language over Windows XP. The haptic device used is
Yr (q, q.q r , qr ) K ds s K L tanh( s )ds (23) PHANToM premium [11], provide 3 degrees of freedom
positional sensing and 3 degrees of freedom force feed-
Given that Yr ( q , q.q r , q
r ) is upper bounded, we back, and connects to the PC via the parallel port (EPP)
propose modified control law as: interface.
5.2 Phat Planning For Guidance

K ds s K L tanh( s )ds (24)
The patient is guided in the real world by the end effector
5 EXPERIMENTAL PLATAFORM FOR DIAG- of PHANToM 1.0, in order to solve a maze constructed of
NOSIS AND REHABILITATION wood, the guidance in the maze has a duration is 23.5
The purpose of this paper is to generate the experimental seconds. In the Fig. 2 shows the workspace where they are
platform, which permit an effective medical rehabilitation
in the shortest time possible, also provide information for
a medical diagnosis. The experimental platform consists of
the patient with physical disability doing recovery exercis-
es, these exercises are based on the solution maze. The
desired trajectory is programmed into the device PHAN-
ToM hptico premium 1.0 [11], this device guides the pa-
tient on the maze.
The patient is stimulated in a kinesthetic way (related
to muscle and tendons) by the haptic device, however the
environment with which interacts is real, the labyrinth is
made of wood but the desired trajectory is programmed
into th device.
In the Fig.1 presents the haptic guidance system, where
X r is the reference position of the desired trajectory, X r
determines the reference velocity of the trajectory,
previously evaluated under the best conditions for
Fig. 2. Experimental platform
interaction. Because the design control signal was
IJSER 2011
5
ISSN 2229-5518
held on the rehabilitation of patients.

The polynomials that represent the trajectories of the
solution maze (see Fig. 3) are based on equations; they This way we perform the task planning considering
represent the position, velocity and acceleration of the convergence in finite time (see Fig. 3). The characteristics
trajectory. The polynomial of the position is as follows, that have the polynomials design allows a patient to be
guided at the beginning of the exercise: smoothly increas-
ing velocity and decreasing when the trajectory is about to
(t t 0 ) 3 (t t 0 ) 4 (t t 0 ) 5
(t ) a3 a 4 a 5
(25) end. In Table I is presented the start and the end time for
(t b t 0 ) 3 (tb t0 ) 4 (tb t0 ) 5 each trajectory represent in the Fig.4.
TABLE 1
INITIAL AND CONVERGENCE TIME OF THE TASKS PRESENTED IN
FIG. 4
Fig. 3. Performance of polynomials and .
The polynomial (t ) corresponds to a smooth trajecto-

ry to the start and end time, which breaks with the inertial
effects due to the state of repose and movement (see Fig. 3),
t is the time, t 0 represents the initial time and tb deter- 5.3 The Inertial Dynamics Effects
mines at the time of convergence. To obtain the velocity is Convergence in finite time by the position and velocity
necessary to derive the polynomial (t ) with respect to r , this benefits the
trajectory is reflected in accelerating q
time, and the acceleration has to derive the polynomial task planning. Fig.5 shows the inertial torque
twice with respect to time. To determine the values of a3 ; H D(q)q using the patient acceleration generated in
a4 and a6 is done as follows, with the conditions the exercise.
(t 0 ) 0 , (t0 ) 1 , (t ) 0 , (tb ) 0 and
(0.5tb ) 0 , so solving the equations and is obtained
that a3 10 , a4 15 and a5 6 .
Fig. 5. Inertial torques one and trhee

Fig. 4. Path planning
IJSER 2011
6
ISSN 2229-5518
However D(q) corresponds to the matrix of inertia of techniques to analyze stability is through the behavior of
the haptic device. In P0, P1, P2, P3, P4, P5, P6, P7, P8, P9, the Lyapunov function (10) and their respective deriva-
P10, P11, P12 and P13 inertia is minimal due to the benefit tive (20).
of the task planning, and the design of the control law
that allows stable tracking [16].
6.1 The Experiments
Rehabilitation process consists on patient performance of
ten exercises, which have different characteristics under
certain parameters. Here are the characteristics of each
exercise:
1. Exercise 0.-The patient is guided to follow up
the trajectory, in this exercise haptic device ac-
tuators are zero, and this way the patient vo-
luntarily moves the PHANToM end effector
(see Fig.6).
Fig. 7. Exercise 1 (with control) . The patient is guided for 10 times

continuously. This allows for success in motor performance
Fig. 6. Exercise 0 (uncontrolled).The patient's performance in the

first experiment without control. The patient is not guided by the
haptic device PHANToM.
2. Exercise 1-10.-The patient is guided by PHAN-

ToM considering the control signal in (24). The
constants have the following value K ds 0.03 Fig. 8. Exercise 1 (with control) . The patient is guided for 10 times
and K L 0.009 . continuously. This allows for success in motor performance
3. Exercise 11.-The patient performs the last exer-
cise with the same features as the first, without If the candidate Lyapunov function is defined postive
control in the actuators (see Fig.8). The patient in every instant of time and also the derivative with re-
is free movement during the experiment. He spect to time of the Lyapunov function is negative semi-
controls his movements. definite then the system is stable. In order to know if the
The Fig.7 shows the performance PHANToM 1.0 work- system (human machine interaction) was stable it is ne-
space, the patient has a certain disturbance index during cessary to consider the behavior of the functions de-
the haptic guidance. Finally, Fig.8 illustrates the perfor- scribed above.
mance of the last exercise of patient, this to control the In Fig 9 shows the behavior of the Lyapunov function
actuators. of Exercise 1. The function is always positive definite,
therefore the system is stable with the patient in the con-
6.2 Stability Analysis trol loop.
On the other hand the derivative of the Lyapunov
The nonlinear control design described in 24 is a ro-
function is shown in Fig 10. The function is negative se-
bust control that allows a stable man-machine interaction.
midefinite at all times, so that the system is stable at all
Control design is based on Lyapunov theory (second
times. Disturbances of the patient during the interaction
method) and passivity. It is well known that one of the
are reflected in the impulses that show the functions.
IJSER 2011
7
ISSN 2229-5518
Fig. 9. The graph that describes the behavior of the Lyapunov

function permirte note that the function was always positive definite, Fig. 11. The interference generated by the patient is reflected in
instantaneous changes of the control signal in the three actuators.
but at certain moments presented impulses, this is due to distur-
bances of the patient
6.3 Discusions
In the Fig.12 and Fig.13 present the graphs representing
the rehabilitation of the patient during each of the 10 exer-
cises. In each of the experiments were performed,
measuring the disturbance index that the patient generated
during each exercise and then assess the performance and
learning via the control signals (see Fig. 9A). The distur-
bance of the patient influences the dynamic error signals S,
in each exercise is calculated
S and is plotted as shown
in Fig. 13.
Fig. 10. The high disturbances generated by the patient did not
prevent the system was unstable.
Considering the forces on individual fingers, which

should be under 30 to 50N in total and the average force
that humans can exert on the index finger is 7N, 6N mid-
dle finger and ring finger without experiencing 4.5N dis-
comfort or fatigue. It is important that the control signal
in the actuators of the haptic device does not exceed the
force of 7N. Thus patient safety prevails in closed loop
and increasing the useful life of the device. In order to Fig. 12. The performance of the patient in the rehabilitation process
ensure patient safety is necessary to observe the behavior is presented in this graphic.
of the nonlinear control law.
In Fig 11 shows the control signal of the three actuators Based on the data obtained, the comparison between the
which does not exceed 7N. Because the design of the con- initial results (exercise 1) and the end (exercise 10) results,
trol law is based on theories that allow for energy analysis shows that the patient acquires the ability to solve the maze
and pasivisar the system with the patient in the control and also that her motor movements are recovered gradual-
loops. ly as the number of exercises is increased.
IJSER 2011
8
ISSN 2229-5518
habilitation. vol. 83, pp. 952-959, 2002.

[4] H.I. Krebs, J.J. Palazzolo, L. Dipietro, M. Ferraro, J. Krol, K.
Rannekleiv, B.T. Volpe, and N. Hogan, Rehabilitation Robot-
ics: Performance-Based Progressive Robot-Assisted Therapy,
Autonomous Robots, vol. 15, no. 1, pp. 7-20, 2003.
[5] W.S. Harwin, J. Patton, and V.R. Edgerton, Challenges and
Opportunities for Robot Mediated Neurorehabilitation, Pro-
ceedings of the IEEE Special issue on medical robotics, vol. 94, no. 9,
pp. 1717 1726, September 2006.
[6] MJ, Eriksson J, Feil-Seifer DJ, Winstein CJ, Socially assistive
robotics for post-stroke rehabilitation, J Neuroeng Rehabilitation,
2007, 4:5.
[7] P. Metherall, EA. Dymond, N. Gravill, Posture control using
electrical stimulation biofeedback: a pilot study, J Med Eng
Technol 1996, 20:53-59.
[8] D. Jack, R. Boian, AS. Merians, M. Tremaine, GC. Burdea, SV.
Adamovich, M. Recce, Z. Poizner, Virtual reality-enhanced
Fig. 13. The absolute value of the error spread in the patient's
stroke rehabilitation, IEEE Trans Neural Syst Rehabil Eng, 2001,
rehabilitation is important. Reflects the progress of the patient at all
times. The patient tends to ideal behavior. 9:308-318.
[9] D. Feygin, M. Keehner and F. Tendick, Haptic Guidance:
Experimental Evaluation of a Haptic Training Method for a
4 CONCLUSIONS Perceptual Motor Skill,Proceedings of the 10th Symposium on
Haptic Interfaces for Virtual Environment and Teleoperator Systems,
The experimental platform design presented in this
2002. HAPTICS 2002, pp 40-47, Orlando, FL, 2002.
work demonstrates the importance of using the technique
[10] F. Clark and K. Horch, Kinesthesia: in Handbook of Percep-
of haptic guidance for rehabilitation and medical diagnosis.
tion and Human Performance, v.1 Sensory Processes and Percep-
One of the factors that influence the experimental platform
tion, New York: Wiley, 1986.
for the stability of the system is the design of the control
[11] SensAble Technologies, Inc., PHANTOM PREMIUM 1.0/1.5A,
law, which promotes stability and robustness in human-
3D Touch Components: Hardware Installation and Technical Ma-
machine interaction, this allows the patient tracking task.
This will take care of the patients physical integration and nual, Revision 6.5, 18 August,2000.
[12] V. S. Rafael Kelly. Control de Movimiento de Robots Manipu-
partial or total damage of the device. Another relevant
ladores. Prentice-Hall, pp 15-37, 2003.
factor is the high performance that has the haptic device,
[13] P. J. N. Romeo Ortega, Antonio Lora and H. Sira- Ramrez.
which in this case is very high due to the high resolution
Pasivity-based Control of Euler-Lagrange Systems. Springer-
optical encoders as well as the under backlash that presents
Verlag, 1998.
PHANToM.
[14] A. Jarillo-Silva and O.A. Domnguez-Ramrez, Joint Control
The patient, gradually rehabilitated, using haptic guid-
Strategy for Haptic Guidance, 2010 Electronics, Robotics and Au-
ance platform leads to reduction in motor rehabilitation
tomotive Mechanics Conference,pp 411-416, 2010.
time, which reduces the cost that this implies. Data pro-
[15] A. Jarillo Silva and O.A. Domnguez Ramrez, Diseo e Im-
vided by the platform are evaluated by specialists to de-
plementacin de Tcnicas de Control No Lineal Basadas en Pa-
termine a diagnosis based on the performance of the pa-
sividad Aplicadas a Guiado Hptico, Master, Research Center
tient.
on Information Technologies and Systems, Autonomous Uni-
versity of Hidalgo State, Hidalgo, Mxico, 2009.
REFERENCES
[16] A. Jarillo-Silva, O.A Domnguez-Ramrez and V. Parra-Vega,
[1] T. Truelsen, B. Piechowski-Jwiak, R. Bonita, C. Mathers, J. Haptic Training Method for a Therapy on Upper Limb, 3rd
Bogousslavsky, and G. Boysen, Stroke Incidence and Preva- International Conference on Biomedical Engineering and Informatics
lence in Europe: a Review of Available Data, European journal (BMEI 2010), pp. 1750-1754, 2010.
of neurology, vol.13(6), pp. 581-598, June 2006.
[2] R. Colombo, F. Pisano, S. Micera, A. Mazzone, C. Delconte, M.
C. Carrozza, P. Dario, and G. Minuco, Robotic Techniques for
Upper Limb Evaluation and Rehabilitation of Stroke Patients,
IEEE Transactions on Neural Systems and Rehabilitation Engineer-
ing, vol.13, no. 3, pp.313- 324, September 2005.
[3] P. S. Lum, C. G. Burger, P. C. Shor, M. Majmundar, and M. van
der Loos, Robot-Assisted Movement Training Compared with
Conventional Therapy Techniques for the Rehabilitation of Up-
per Limb Motor Function After Stroke,Arch. Phys. Meical. Re-
IJSER 2011
9
ISSN 2229-5518
Prevalence of Eating Disorders among

Female Students of Tonekabon University
(Tonekabon-Iran)
Roxanna Ast , Azade Fadavi Roodsari
Abstractaim of this research is to study the prevalence of eating disorder amongst female students of Tonekabon
University. The community being studied is the female students of Tonekabon University.300 students were randomly
selected and requested to complete the eating attitude test-26.
Index Terms Eating disorders, Female students, Tonekabon University

1 INTRODUCTION
E ating disorder presents a serious problem these days.

Every year the number of people suffering from ano-
rexia and bulimia increase and consequents of the
then purge themselves of what they have eaten usually by
forcing themselves to vomit and sometimes by means of
laxatives, diet pills, diuretic pills or excessive exercising.
disease may be health or even life threatening.[1] These behaviors occur at least several times per week for
This disorder presents a significant problem among ado- months on end, the condition is usually related to over
lescent and young women in many westernized countries concern with ones weight and shape, and is accompanied
and is associated with nervous, physical and psychiatric by feeling of shame, disgust and being out of control.[8]
problems.[2] Like anorexia nervosa, Bulimia was recognized to occur
Anorexia nervosa is a psychological and physical condi- as early as the 17th century. [9]
tion of semi starvation in which individuals weigh 85% or According to Wilson [10]to be diagnosed with bulimia
less of what would ordinary be there healthy body nervosa individuals must experience episodes of binge-
weight, resulting in physical impairments and in the 90% eating at least twice a week on average, for three to six
of patients who are females Cessation of menses. months.
This condition is to due to highly restricted food intake, In addition to the primary eating disorders, several other
often accompanied by excessive exercise and sometimes conditions occur among individuals with psychiatric dis-
purging by self-induced vomiting, laxative use or other orders that may markedly affect eating behavior and
means. weight. For instance, individuals with severe depression
These behaviors are usually related to obsessional and a experience an increase in appetite and food cravings. Pa-
perfectionist thinking that focuses on a distorted body tients with psychic delusions due to Schizophrenia or
image and fear of becoming fat.[3] other conditions may think food is poisoned and refuse to
Feeling normal fullness after eating is felt as discomfort eat.[11]
and experienced as a failure of control, moral weakness Recent studies have found that EDNOS (Eating Disorder
and a source of great guilt. These perfectionists are failing Not Otherwise Specified) is the most common eating dis-
in their major (anorexia) project and will redouble their order, diagnosis both in outpatients and inpatients set-
effort by eating nothing for a day or even restricting their tings. Underweight patients that do not report over-
water intake. This can precipitate death through cardiac evaluation of shape and weight are a distinctive and
arrest, particularly if they are also vomiting the little in- scarcely studied subgroup of EDNOS.
take they do allow.[4] Their self-evaluation is largely or exclusively based on
The association between anorexia nervosa and depression their abilities to control their eating purge.[12]
has long been recognized by clinicians.[5] The incidence of eating disorders in females has been ex-
A study in Sweden using structured interview for DSM- tensively studied in both anorexia nervosa and bulimia
lll-R criteria found that 85% of patients with anorexia nervosa.[13]
nervosa (AN) had a depressive disorder.[6] In a study of 105 patients with eating disorders, Braun
About half of sufferers eventually develop binge-eating [14] found that the life time prevalence of any offensive
episodes that is, periodic decontrol over eating on inca- disorder was 41.2 % in anorectic restrictors, 82% in ano-
pacity to satiate. [7] rectic bulimics, 64.5% in patients with bulimia nervosa
and 78% in patients with bulimia nervosa with a past his-
Bulimia nervosa (BN) is a condition in which individuals tory of anorexia nervosa.
binge, eat large calories, up to 2000 at a time or more and According to Treasure [15] the present time prevalence of
IJSER 2011
10
ISSN 2229-5518
all eating disorders is about 5%.cultural, social and inter- orders.

personal elements can trigger onset and change in net-
works can sustain the illness. TABLE 1. EAT-26 TEST RESULTS
Although it is clear that anorexia nervosa occurs in men SAMPLE POP- STUDENTS PERCENTAGE
as well as women, and in younger as well as in older ULATION SCORING OF STUDENTS
people, few studies report incidence rate for males or for
ABOVE WITH A SCORE
people beyond the age of 35.
The majority of male incidence rate reported was below 20 HIGHER THAN
0.5 per 100,000 populations per year. [16] 20
Studies have reported the female to male ratio to be
around 11 to 1.[17] 300 61 20.33 %
On an overall female rate of 15.0 per 100,000 population STUDENTS
per year, Lucas [18] reported a rate of 9.5 for 30-39-year-
old women,5.9 for 40-49-year-old women,1.8 for 50-59-
This amount forms 20.33 % of the total population of fe-
year-old and 0.0 for women aged 60 and over .
male students at Azad university of Tonekabon.
According to a research by Casper [19], women who had
recovered from anorexia nervosa rated higher on risk
avoidance, displayed greater restraint in emotional ex-
pression and initiative, and showed greater conformance
to authority than age-matched normal women.
Lucas [20] found that the age-adjusted individuals rates
of AN in females 15-24-year-old showed a highly signifi-
cant linear increasing trend from 1935-1989,with an esti-
mated rate of increase of 1.03 per 100.000 person per ca-
lendar year.
2 METHOD
This is a descriptive study and aims to objectively de-
scribe the features of the desired community.
The research took place over a period of 4 months.
(March 2010 - June 2010). Fig. 1.
In this research, based on the total population -6000 fe-
male students-a sample of 300 were randomly selected.
Subjects were students who attended classes at university
throughout the week. The team, attended classes on dif- Among the three sub-scales namely: Oral Control food
ferent hours of different days and selected few names habits and Esurience (desire to eat), the most points be-
from the participants of the class-regardless of their BMI, longed to food habit sub-scale (86.88%) second is oral con-
body shape and weight-afterwards, subjects completed trol and the rest were high in esurience sub-scale.
the questionnaire. EAT-26 was used as the main method The main problem and barrier throughout the research
of data collection. was the subjects orientation towards the questions on the
The eating attitude test (EAT-26) has been proposed as an test which was resolved by the researchers explanation in
objective, self report, measuring the symptoms of anorex- person.
ia nervosa. It has been used as a screening instrument for
detecting previously undiagnosed cases of anorexia ner- 4 DISCUSSION
vosa in populations at high risk of the disorder. [21] The purpose of this study is to measure the prevalence of
Subjects can score between 0 78.any score above 20 is eating disorder in female students of Tonekabon Univer-
diagnosed with eating disorder. sity. According to the results 20.33% of subjects were
The EAT-26 can only show some symptoms of eating dis- identified with eating disorder.
order. To classify the type of disorder (AN or BN) a diag- In our country, in the field of eating disorder, few studies
nostic interview will need to take place. have been carried out.[22]
In measuring the prevalence of eating disorders amongst
high school students (male and female) in Sari- North of
3 RESULTS Iran-In the academic year of 2002-2003 in 10.5% of the
The results of this study showed that among 300 cases students, abnormal attitude towards eating was observed.
under review, 61 of them received a score higher than 20 This conclusion suggests the frequency of abnormal atti-
on the EAT-26 test and were diagnosed with eating dis- tude toward eating disorder is more or less similar to that
IJSER 2011
11
ISSN 2229-5518
of other communities. ethnic minority women: a review of the emerging litera-

Patients with this attitude most likely have some psycho- ture, Current psychiatry review,2009, vol.5,issue.3, p 182-
logical disorder such as, depression. If this issue is not 193
managed in the appropriate manner, many side effects [3],[8],[11] Yager J., Encyclopaedia of psychotherapy, eat-
will follow. [23] ing disorder, 2002, p671, 672
AN and BN occur disproportionately often in industria- [4] Churven P., Hunger strike or medical disorder? Is
lized (vs. economically less developed countries)nations, anorexia caught in our flawed dichotomy between body
but are otherwise reported to occur with surprisingly uni- and mind?, ANZJFT , 2008 , Vol.29, Num.4 , pp.184-190
form prevalence in developed parts of Europe, Asia and [5],[7],[9],[24] Maj M., Halmi K., Lopes J.J , Sartorios N.,
the Americas.[24] Eating disorder, 2003, vol.6 , ISBN :0-470-84865-0
Young girls and women have been influenced by western [6] Ivarsson T., Rastam M., Wentz E., Gillberg I., Gill-
culture and have adapted the same behaviour towards berg.C ., depressive disorders in teenagers onset anorexia
issues like appearance and body shape.[25] nervosa : a controlled longitudinal , party community-
The extent of the impact has not, yet, been provided in based study .(2000),compr, Psychiatry , 41 :398-403
statistics.[26] [10] Wilson GT., Sysko R., Frequency of binge eating epi-
In a study to determine the prevalence of eating disorder sodes in bulimia nervosa and binge-eating disorder: di-
among female high school students in Tehran-Capital of agnostic consideration, International journal of eating
Iran-results show that 9.0 % were diagnosed with anorex- Disorder, Nov 2009, 42(7): 603-10 ISSN: 0276-3478
ia, 23.3% with Bulimia and 63.7% with a mild eating dis- [12] Grave R., Calugi S., Marchesini., Underweight eating
order.[27] disorder without over-evaluation of shape and weight:
In another study by Seyedi [28] the prevalence of eating atypical anorexia nervosa? , International journal of eat-
disorder among 600 female high school students in Ker- ing disorders, 2008, 41: 8705-712
man was found to be 7.6%. [13],[29] Button E., Aldridge S., Palmer R., International
Some studies suggest an increase in incidence of AN over Journal of eating disorders, males assessed by a specia-
various time period, whereas others report an increase in lized adult eating disorders service: patterns over time
incidence followed by stabilization.[29] and comparisons with females ,41:8 ,758-761, 2008
Epidemiological studies show that eating disorder is [14] Braun D.L ., Sunday S.R., Halmi K.A., psychiatric co
mostly a white female disease in industrialized countries. morbidity in patients with eating disorders, 1994 , psycho
Societies, in which, slimness is a main criteria of attrac- , med ,24 : 859-867
tion.[30] [15] Treasure J.,Claudino AM., Zucker N Lancet, Eating
Unfortunately, in the past few years eating disorders has disorder,13 Feb. 2010, vol. 375 pp 583-593,England
increased in Iran. As the most people diagnosed with this [16] Turnbull L.S., Ward A., Treasure J., Jick H., Derby L,.
disorder are the young generation, therefore, education The demand for eating disorder care, an epidemiological
on this matter is one of the most important means of han- study using the general practice research data base, 1996 ,
dling this issue. British journal of psychiatry ,169 , pp 705-712
Generally, the level of research conducted on aetiology
and treatment of eating disorder In Iran is very low. [17] Hoek H.W., Bartelds , A.I.M ., Boosveld J.J.F, Vander
Young girls of today will be the mother of our future gen- Graaf, Limpense Y., V.E.L ., Maidwald M & Spaij C.J.K ,
eration and their mental and physical health plays an im- Impact of urbanization on detection rates of eating dis-
portant role in the future of our children and country. orders, 1995, American journal of psychiatry ,152, pp
Therefore, this field need far more extensive research and 1272-1278
exact aetiology and strategies for prevention and therapy. [18],[20] Lucas A.R, Crowson C.S, OFallen , W.M & Mel-
toy L.J, The ups and downs of anorexia nervosa , 1999 ,
5 CONCLUSION international journal of eating disorders , 26 , 397-405
The results show that, based on the eat-26 test criteria, [19] CASPER RC., personality features of women with
from the total population under study, 61 students, and good outcome from restricting anorexia nervosa, psycho-
equal to 20.33% of the total number of students- were di- somatic medicine, 1990, Vol.52, Issue.2, p156-170,
agnosed with eating disorders. Eating disorders are [21] The eating attitude test, Psychometric features and
breaking out rapidly in our society and among the young clinical correlates, psychological medicine, 1982, 12, p
generation and if not dealt with properly serious conse- 871-878
quences will to be expected. [22] Babaee S., Relationship between core beliefs and abil-
ity to eat with symptoms of eating disorder in students of
REFERENCES: shahid beheshti University, General psychology Senior
[1] Pasternak B., Czekaj J., Sabina level of anxiety con- thesis , 2006.
nected with putting on weight as a predictor of bulimia [23] Zarghami M.,Chyme N., Prevalence of non-natural
and anorexia.,2009 , physiothera vol.17,issue.3, p36 5p. approach to eating and the factors correlated with the
[2] Brown M., Fary Cachelin., Faith A., eating disorder in high school students,2002, Journal of Mazandaran Uni-
IJSER 2011
12
ISSN 2229-5518
versity of medical sciences, Autumn 2003 ; 13 (40) :70-78

[25] Gaffari Nejd A., eating disorders in female students
of kerman university of medical sciences, ministry of
health and medical education, Kerman University of
medical sciences, 1998
[26],[28] Seyedi F., Sanayi F., Prevalence of eating disord-
ers in female students in high schools of Kerman, 2004,
Kerman university school of medical sciences.
[27] Nobakht M., epidemiology study of eating disorders
in second year high school female students in Tehran,
Thesis (MA) 1998, Ferdowsi University of mashhad, Fa-
culty of educational sciences and psychology
[30] Szabo C.P. , Hollands. C ; Abnormal eating attitudes
in secondary-school girls in south Africa; South Africa
medical journal , 1997 , 87(4) , 524- 6: 528-30
IJSER 2011
13

ISSN 2229-5518
Impact of Leverage on Firms Investment

Decision
Franklin John. S, Muthusamy. K
Abstract - The present paper is aimed at analyzing the impact of leverage on firms investment decision of Indian pharmaceutical
companies during the period from 1998 to 2009. To measure the impact of leverage on firms investment decision, pooling
regression, random and fixed effect models are used by taking, leverage, sales, cash flow, Return on Asset, Tobins Q, liquidity and
retained earnings as independent variable and investment as dependent variable. In addition , we demarcate between three types of
firms (i) Small firms, (ii) Medium firms and (iii)Large firms. The results reveal that a significant positive relationship between leverage
and investment, while we found a negative relationship between leverage investment for medium firms and positive relationship
between leverage and investment in large firms. Our econometric results reveal an insignificant relationship between the two
variables for medium and large firms.
Index Terms-- Investment, Tobins Q, Cash flow, Liquidity, ROA, Size and Retained Earnings.
INTRODUCTION decision has a direct influence on the firms asset

structure, more over in their degree of liquidity
I nvestment is a crucial economic activity in the
corporate financial management. Such an

and consists of spending the financial funds for the
purchase of real and financial assets for the firm. In
activity leads to the countrys economic
order to gain cash and the growth of the wealth of
development provide employment to the people
firms owner. The investment decision and the
and to eliminate poverty .This paper investigates
financing decision are interdependent that is the
the effort of debt financing on the firms investment
investment decision is adopted in relation to the
decision on pharmaceutical industry in India. It
level of financing source but the option to invest is
plays a significant role in the countrys economic
also crucial in order to calculate the level of
and industrial development and trade and to
financing capitals and the need for finding their
prevent diseases for increasing the life of people.
sources.
This industry is providing a basic material to other
industrial sectors. It requires capital for financing As far as the hierarchy of financing
firms assets. Among the different sources of fund, sources as it exists in the economic literature, is
debt is a cheaper source because of its lowest cost concerned, cash flow is the cheapest financing
of capital. The investment decision of the firm is of sources followed by debts and in the end, by its
three categories that can be adopted by firms issuing of new shares. Debts can be cheaper than
management besides the financing decision and the issue of new shares because the loan contract
the net profit allocation decision. The investment can be created as to minimize the consequences of
IJSER 2011
14

ISSN 2229-5518
information problem. Giving the fact the degree of Second, several emerging economies, even until
information asymmetry and the agent costs the late 1980s, suffered from financial depression,
depend on the peculiarities of every firm, such with negative real rates of interest as well as high
firms are more sensitive to financial factors than levels of statutory pre-emption. This could have
other. The debt limit of the firms is determined in meant a restricted play of market force for resource
the view, since interest payment is tax deductible, allocating.
the firm prefers debt financing to equity and it

Issues regarding the interaction between
would rather have an infinite amount of debt,
financing constraint and corporate finance have,
However, this leads to negative equity value in
however, gained prominence in recent years,
some status so that the firm would rather go
especially in the context of the fast changing
bankrupt instead of paying its debt. Therefore debt
institutional framework in these countries. Several
to remain risk-free, lenders will limit the amount of
emerging economies have introduced market-
debt. They can limit the debt by accepting the
oriented reforms in the financial sector. More
resale value of capital as collateral and ensuring
importantly the institutional set-up within which
that this value is not lower than the amount of
corporate houses operated in the regulated era has
debt, so that they can recover their money in case
undergone substantial transformation since the
of bankruptcy. Alternatively, lenders may limit the
1990s. The moves towards market-driven
amount of debt in order to ensure that the marker
allocation of resources, coupled with the widening
value of equity is always non-negative and
and deepening of financial market, have provided
bankruptcy is sub-optimal for the firm.
greater scope for corporate house to determine
While there is by now a rapidly expanding their capital structure.
literature on the presence of finance constraints on

The rest of the paper unfolds as follows.
investment decisions of firms for developed
Section II discuses the historical background of the
countries , a limited empirical research has been
study. Section III explains methodology, data,
forthcoming in the context of developing countries
variable description and the data employed in the
for two main reasons. First until recently, the
paper. Section IV presents the results and discusses
corporate sector in emerging markets encountered
robustness check followed by the concluding
several constraints in accessing equity and debt
remarks in the final section.
markets. As a consequence, any research on the
interface between capital structure of firms and THE BACKROUND OF THE STUDY
finance constraints could have been largely
Several authors have studied the impact of
constraint- driven and have less illuminating.
financial leverage on investment. They reached
IJSER 2011
15

ISSN 2229-5518
conflicting conclusions using various approaches. commitments invest less no matter what their
When we talk about investment, it is important to growth opportunities . Theoretically, even if
differentiate between overinvestment and under- leverage creates potential underinvestment

investment. Modigliani and Miller (1958) argued incentives, the effect could be reduced by the firm
that the investment policy of a firm should be corrective measures. Ultimately, leverage is
based only on those factors that would increase the lowered if future growth opportunities are
profitability, cash flow or net worth of a firm. recognized sufficiently early.
Many empirical literatures have challenged the

Another problem which has received much
leverage irrelevance theorem of Modigliani and
attention is the overinvestment theory. It can be
Miller. The irreverence proposition of Modigliani
explained as investment expenditure beyond that
and Miller will be valid only if the perfect market
requires to maintain assets in place and to finance
assumptions underlying their analysis are satisfied
expected new investment in positive NPV projects
.However the corporate world is characterized by
where there is a conflict between manager and
various market imperfections costs, institution
share holder. managers perceive an opportunities
restrictions and asymmetric information. The
to expand the business even if the management
interaction between management, shareholders
under taking poor projects and reducing
and debt holders will generate frictions due to
shareholders welfare .The managers abilities to
agency problems and that may result to under-
carry such a policy are restrained by the
investment or over-investment incentives. As
availability of cash flow and further tightened by
stated earlier one of the main issues in corporate
the financing of debt. Hence, leverage is one
finance is whether financial leverage has any
mechanism for overcoming the overinvestment
effects on investments policies.
problem suggesting a negative relationship
Myers (1977), high leverage overhang reduces the between debt and investment for firm with low
incentives of the shareholder-management growth opportunities. Does debt financing induce

coalition in control of the firm to invest in positive firms to make over-investment or under-
net present value of investment opportunities, investment? The issuance of debt commits a firm to
since the benefits accrue to the bondholders rather pay cash as interest and principal. Managers are
than the shareholders thus ,highly levered firm are forced to service such commitments .too much
less likely to exploit valuable growth opportunities debt also is not considered to be good as it may
as compared to firm with low levels of leverage a lead to financial distress and agency problems.
related under investment theory centers on a

Hite (1977) demonstrates a positive
liquidity affect in that firm with large debt
relationship because given the level of financial
IJSER 2011
16

ISSN 2229-5518
leverage an investment increase would lower investment and sell off unprofitable business
financial risk and hence the cost of bond financing. divisions.
In contrast Deangels and Masulis (1980) claim a

Daddon and Senbets (1988) hypothesis on
negative relationship since the tax benefit of debt
the relationship between bond financing and
would compete with the tax benefit of capital
capital investment which is conditional on from
investment. Dotan and Ravid (1988) also show a
specific variables such as tax shield, retention
negative relationship because investment increase
ability, capital intensity and insider equity
would raise financial risk and hence the cost of
ownership. Josephic Kang(1995)who found that the
bond financing how the investment increase affects
level of bond financing has negative relationship
financial risk and the sub suitability between tax
with level of investment.
shields and hence; financial leverage may depend
on firm-specific factors. Whited (1992) has shown how investment
is more sensitive to cash flow in firms with high
Jensen (1986) points out that liabilities can
leverages as compared to firms with low leverage.
help avoid overinvestment by reducing the cash
Cantor (1990) showed that investment is more
flow left up to corporate managers own discretion
sensitive to earnings for highly levered firms
and constraining investment in investment
projects that might be desirable for corporate Mc connell and Servaes (1995) have examined a
mangers but not desirable for companies future large sample of non financial United State firms for
profitability. Jensen argues that whether liabilities the years 1976, 1986 and 1988. They showed that
restrain overinvestment depends largely on for high growth firms the relationship between
whether companies have growth opportunities. In corporate value and leverage is negatively
short, Jensen points out those liabilities have not correlated. Also the allocation of equity ownership
only the negative effects of restraining between corporate insiders and other types of
overinvestment by low-growth companies. Like investors is more important in low growth than in
Jensen(1986),Stulz(1990) and Hart and Moore(1995) high growth firms.
argue that liabilities effectively restrain
McConnelll and Servaes (1995) use cross-sectional
overinvestment. They reason that increased
data to analyze U.S listed companies in 1976, 1986
liabilities, by enlarging repayment obligations, not
and 1988, and find two faces of debt, meaning
only curtail free cash flow but also raise the
that enterprise value was negatively correlated
possibility of corporate bankruptcies, thus
with the debt ratio of companies with high growth
prompting corporate managers to reduce
opportunities. Lang et al. (1996), based on an
analysis of the relationship between the debt ratio

IJSER 2011
17

ISSN 2229-5518
and the rate of growth of companies, point out that focused counterparts and diversified companies
for companies with fewer investment make larger investments (net cost of capital/sales)
opportunities (i.e. companies with a low Tobins than focused counterparts. They also point out that
Q), there is a negative correlation between the debt debt ratio influence management decisions on
ratio and the investment. The estimation results investment and that diversified companies can
from their studies do not find a negative overcome debt ratios through the distribution of
correlation between the debt ratio and the growth liabilities by corporate managers.
rate for companies with abundant growth

Arikawa et al,(2003) adopt the method of
opportunities. In other words, for companies with
estimation used by Lang et al.(1996) and point out
investment opportunities, increased liabilities do
that the main bank system in Japan helped amplify
not necessarily hamper growth.
the disciplinary function of liabilities, particularly
Lang et al (1996) found that there is negative for low-growth companies.
relation between leverage and future growth at the

Aivazian et al (2005) analyses the impact of
firm level and for diversified firms, at the business
leverage on investment on Canadian industrial
segment level. Also debt financing does not reduce
companies cover the period from 1982 to 1999.
growth for firms known to have good investment
They found a negative relationship between
opportunities, but it is negatively related to the
investment and leverage and that the relationship
growth for firms whose growth is not recognized
is higher for low growth firms rather than high
by the capital market.
growth firms.
Myers (1997) has examined possible difficulties

Ahn et al (2006) found that diversified companies
that firms may face in raising finance to
tend to have higher debt ratios then focused
materializing positive net present value (NPV)
counter parts and diversified companies make
projects, if they are highly geared. Therefore, high
larger investments than focused counter parts.
leverage may result is liquidity problem and can
They also point out that debt ratio influence
affect a firms ability to finance growth. Under this
management decisions on investments and that
situation, debt overhang can contribute to the
diversified companies can overcome the
under-investment problem of debt financing. That
constraints of debt ratio through the distribution of
is for firms with growth opportunities debt have a
liabilities by corporate managers.
negative impact on the value of the firm.
Mohanprasadsing Odit and Chitto (2008) analyze

Ahn et al.(2000), found that diversified
the impact of leverage on firms investment on 27
companies tend to have higher debt ratio than
maturation firms that are quoted on the stack
IJSER 2011
18

ISSN 2229-5518
exchange Mauritians for the year 1990 04. They the profitability of the firm i; LIQiyt-1 represents
found that leverage has a significant negative effect liquidity of firm i : RETESi,t-1 is the retained
on investments, Suggesting that capital structure earnings of the firm i .
plays an important role in the firms investment

The data used in this paper are from the annual
policies while the negative relationship persist for
report of Indian pharmaceutical companies which
low growth firm, this is not the case for high
are listed in Bombay stock exchange and this data
growth firm.
have been collected from CIME browse data base
Thus the previous studies have verified the impact of top 25 companies based on sales from the period
of leverage on firms investment decision as well as 1998 2009
the effect of leverage in restraining over

LEV: Lev denotes leverage. We have used the same
investment and facilitating under investment.
definition of leverage as lang. et al (1996), namely
These studies suggest that leverage restraining
the ratio of total liabilities to the book value of total
over investment but likely cost under investment.
assets. He pointed that the market value leverage
Thus in this paper an attempt is made to more
gives too much weight to the deviations in equity
clearly the leverage impact on firms investment
value. The book value of leverage does not reflect
decision on pharmaceutical companies in India.
recent deviation in the market valuation of the
METHODOLOGY DATA AND VARIABLES firm. If leverage has a significant negative effect on
DESCRIPTION investment, two interpretations can be adopted.
First, it would mean that capital structure plays an
We estimate a reduced form of investment
important role in the firms investment policies;
equation to examine the effect of leverage on
second, it can also be explained by an agency
investment the specification is similar to Aivazian
problem between the agents and the shareholders.
Ge and Qiu (2005).This is as follows:
If managers are overburdened by debt they may
Ii t/Ki, t-1= + [CFit/Ki, t-1] +1Qi, t-1 + 2 give up projects which may yield positive net
LEVi,t-1 +3SALEi,t-1+4ROAi,t-1+5LIQiyt- present values. Also there will be support for both
1+6RETESi,t-1+ i,t the underinvestment and overinvestment theory.
Where Ii t represents the net investment of firm i TOBINS Q: we use prefect and wiles (1994) simple
during the period t; Ki, t-1is the net fixed asset; Q (market value + liabilities / book value of assets)
CFit is t5he cash flow of firm i time t: Qi, t-1is the as a proxy for growth opportunities defined as the
Tobins Q: LEVi,t-1 represents the leverage: market value of total assets of the firm divided by
SALEi,t-1 stands for net sales of firm i ; ROAi,t-1 is the book value of assets. Market value of the firm
IJSER 2011
19

ISSN 2229-5518
is the sum of total liabilities, the value of equity projects. The purpose of allocating money to
shares and the estimated value of preference project is to generate a cash inflow in the future,
shares. The market value of preference share is significantly greater than the amount invested.
calculated as preference divided multiply by ten That is the objective of investment is to create
which measures growth opportunities and it, shareholders wealth. In order to eliminate any size
compare the value of a company given by financial effect. We normalize this measure by taking the
market with the value of a company, Tobins q book value of assets; this method was utilized by
would be 1.0 if Tobin q is greater than 1.0 then the Lehn and Poulson (1989) and Lan et al (1991).
market value is greater than the value of the

PROFITABILITY (ROA): Profitability is measured
companies record assets. This suggests that the
in terms of the relationship between net profits and
market value reflects some unmeasured or
assets. It is calculated as earning after tax adds
unrecorded assets of the company. High Tobins q
interest minus tax advantage on interest divided
values encourage companies to invest more in
total fixed assets. It shows the operating efficiency
capital because they are worth more than the
of the total funds over investment of a firm and is
price they paid for them. On the other hand, if
another important variable that are utilized to
Tobins q is less than 1, the market value is less
growth opportunities as it tries to explain how
than the recorded value of the assets of the
much the assets that the firm is employing in
company.
contributing to the total profitability. Therefore
SALE: sale is measured as net sales deflated by net investment in assets contributes to the profitability
fixed assets. Which measures the efficiency with and we can proxy high profitability with high
net fixed assets is measured. A high ratio indicates growth firms.
a high degree of efficiency in asset utilization and a

LIQUIDITY (LIQ): the liquidity ratio is measured
low ratio reflects inefficient use of assets.
by the current assets divided by the current
CASH FLOW: Cash flow is measured as the total liability and is the ability of firms to meet its
of earning before extraordinary items and current obligations. Firms should ensure that they
depreciation and is an important determinant for do not suffer from lack of liquidity as this may
growth opportunities. If firms have enough cash result in to a sate of financial distress ultimately
inflows it can be utilized in investing activities. It leading to bankruptcy. Lack of liquidity can lead to
also provides evidence that investment is related to a struggle in term of current obligations, which can
the availability of internal funds. Cash flow may be affect firms credit worthiness, Bernake and Gerler
termed as the amount of money in excess of that (1990) argued that both the quantity of
needed to finance all positive net present value of investment spending and its expected return will
IJSER 2011
20

ISSN 2229-5518
be sensitive to the credit worthiness of borrowers. performed with a large chi-square values indicate
That leads us to say that investment decisions of of low P-value. We reject that the pooled estimate
firms are sensitive to current liquidity. However, is appropriate. The second to compare random
Firms with high liquidity give the signal that funds effect estimate with fixed effect estimate, the
are tied up in the current assets. Hausman test is performed. If the model is
correctly specified and it the effect are uncorrelated

RETAINED EARNINGS (RETES): represents the
with independent variables the fixed effect and
amount of business savings meant for bloughing
random effect should not be different a high chi-
back. These are the most favored sources of finance
square value is indicate of appropriateness of the
for corporate firms. There is a significant difference
fixed effect.
in the use of internally generated funds by the
highly profitable corporate relative to the low

profitable firms
RESULT OF SMALL FIRMS
RESULT AND DISCUSSION:

Table 1 brings out the regression result of small
This section portrays the result from the firms. It shows that the leverage has a positive
regression estimation, we present result for the impact on investment at the 5% significant level.
small size, medium size and larger sized firm is The impacts of other variables on have the
classified based on the size. The smaller size is expected signs. The retained earnings have a
obtained by subtracting mean from standard significant positive impact on investment. To
deviation of total asset and larger size is obtained identify which empirical methodology pooling
by adding mean value of asset to standard random effect or fixed effect regression is most
deviation. The median sized firms are those firms suitable, we perform two statistical test the first the
which are not belong to both categories of the firm. Lagrangian Multiplier (LM) test of the random
The econometric result for the sample firms is effect model. The null hypothesis is that individual
showed the pooled estimates; random effect effect ui is 0. The chi-square value is 25.74 thus the
estimates and fixed effect estimates on the T values null hypothesis is rejected at 1% level of
are shown in the parenthesis. Two statistics are significance. The results suggest that the rho effect
used in order to identify, which methodology is is not zero and the pooling regression is not
appropriate to establish the relationship between suitable in this case the regression co-efficient
leverage and investment. First we compare the leverage on small firms from the pooling
pooled estimates and random effect estimates. The regression is equal to 1.3451 and is not significant.
second Lagrangian Multiplier (LM) test is The regression co-efficient of leverage of firms
from
IJSER 2011
21

ISSN 2229-5518
Random effect and fixed effect model are 3.4868 effect and fixed effect models suggesting that
and 1.8200 respectively. The regression ignoring individual firm effects leads to an
co-efficient from the polling regression are much underestimation of the impact of leverage on
smaller than those estimated from the random investment.
Table 1:-Regression result of small firms
We conduct the Hausman specification test hypothesis is rejected at the 1% significance level.
to compare the fixed effect and the random effect The results suggest that the fixed effect model is most
models .If the model is correctly specified and of appropriate in estimating the investment equation.
individual effects are uncorrelated with independent
Leverage is statistically significant at 1%
variables the statistics are showed that the null
and 5% level of significant and is positively related to
IJSER 2011
22

ISSN 2229-5518
investment. A 1 unit increase in the leverage leads to worthiness loss of creditors confidence and this is not
an increase by 3.4568 units in investments this the case as shown by the results from the above table.
implies that a leverage increases in small firms is also
From the table it is observed that TobinQ is
increase a investment of firms because firms do not
negatively related with investments and not
have a adequate asset cushion for financing the
statistically significant.
projects. Thus, in a small sized firm tend to because
more dependent on debt as a source of finance to RESULT OF MEDIUM FIRMS
finance the projects.
Table No 2 Reveals that the regression
The table also reveals that small firms are results of medium firms. The calculated f value is
under utilizing their fixed assets and it would affect greater than table value. Hence the selected variables
the ability in generating the volume of sales and the are significantly associated with investment during
co-efficient value is -0.001 and it is not statistically the period. Further it shows that the leverage has no
significant. impact on investment in medium firm but it has
negative relationship with investment during the
The co efficient value of ROA is 0.0003
period of study. In order to identify which
and is not statistically significant but positively
methodology-pooling random effect or fixed effect
related with investment. It indicates the operating
regression model is most suitable, we perform two
efficiency of the employed funds over investment is
statistical tests, the first the LM test of the random
positive. Higher the ROA is also attracting funds
effect model. The null hypothesis is that individual
from investors for expansion and growth.
effect ui is o. The chi-square value is 4.15. Thus the
Cash flow and retained earnings are null hypothesis is rejected at 1% level of significance.
positively related with investments not statistically The results suggest that the rho effect is zero and the
significant and coefficient value is 0.2264 and 0.0020 pooling regression is suitable in this case. The
respectively. This implies that the issuance of debt regression the co efficient of leverage on medium
engages the firm to pay cash as interest and principal firms from the pooling regression equal to 1.6543 and
with availability of free cash flow and internally is not significant. The regression co-efficient on
generated funds. leverage from random and fixed effect model-0.7797

and-1.6543 respectively. The regression co-efficient
Liquidity is negatively related with from the pooling regression are greater than the those
investments and is not statistically significant and the estimated from the random and fixed effect model
regression co-efficient value is 0.01667. It implies suggesting that the individual effect of a firm leads to
that the failure of a firm to meet its obligation due to an estimation of the impact of leverage on
lack of sufficient liquidity will result in poor credit investment.
Table 2:-Regression result of medium firms
IJSER 2011
23

ISSN 2229-5518
We conduct the Hausman specification test statistics are reported that the fixed effect model is
to compare the fixed effect and random effect most appropriate in estimating the investment
models. If the model is correctly specified and if equation because the R2 value of fixed effect model is
individual effects are uncorrelated with independent greater than random effect model.
variable, the fixed effect and random effect estimates
should not be statistically different. Further these
IJSER 2011
24

ISSN 2229-5518
Leverage is not statistically significant at 1% regression model is most suitable. we perform two
and five per cent level of significance and is statistical test, the first the LM test of the random
negatively related with investment. This implies that effect model. The null hypothesis is that individual
leverage has no impact in medium firms investment effect ui is 0. The chi-square value is 2.26. Thus null
decision. It is because of inadequate cash flow and hypothesis is rejected @ 1% level of significance.
ploughing back of funds. Hence medium sized firms The results suggest that the rho effect is not zero and
are making investment decision based on the internal the pooling regression is not suitable in this case. The
financial resources. The table further reveals that the regression co-efficient of leverage on large firms
medium firms are under utilizing there fixed assets from the pooling regression is equal to 23.7516 and is
and it would effects the ability in generating the not significant. The regression co-efficient on
volume of sales and coefficient value is -0.0016 and leverage from random effect and fixed effect model
is not statistically significant. The co efficient value 9.5758 and 23.7516 respectively. The regression co-
of ROA is -0.0012 and is not statistically significant efficient from the pooling regression Model is greater
but negatively related with investment. The cash than those estimated from the random and fixed
flow and Retained associated with in order earnings effect model suggesting that the individual effect of a
are positively associated with investment and they are firm leads to an estimation of the impact of leverage
statistically significant at 1% and 5% level of on investment.
significant with investment. It indicates that higher
We conduct the Hausman specification test to
the cash flow and retained funds higher will be the
compare the fixed effect and random effect model if
investment. Liquidity and Tobinq are not
the model is correctly specified and if individual
statistically significant with investment the TobinQ
effect are an correlated with independent variable the
also requested firm value and hence may be affected
fixed and random effect are un correlated with
by leverage. But proxies in this do not Influence the
independent variable the, fixed and random effect
investment because the leverage has no impact on
estimate should not be statistically different further
investment in medium firms.
these model is most appropriate that the fixed effect
RESULT OF LARGE FIRMS model is most appropriate in estimating the
investment equation because the R2 value of fixed
Table No 3 Shows that the regression results of large
effect model is greater than the random effect model.
firms. The calculated f value is greater than table
value. Hence the selected variables significantly The table also revels that the co-efficient
associated with investments during the period of value of variables like sales, ROA and TobinQ are
study. Further it shows that the leverage has no negatively related with investment and also they are
impact on investment In large firms but it has not significant in the leverage firms.
positive relationship with investments during the
Cash flow and Retained earnings are
period of study. In order to identify which
positively associated with investment in large firms
methodology-polling, random effect fixed effect
and are statistically significant it is because of heavy
IJSER 2011
25

ISSN 2229-5518
demand for its product in national and international investment. We conclude that the leverage is not
market. Liquidity is negatively related with influenced the investment decisions in large sized
investment and is not statistically significant with pharmaceutical firms in India.
Table 3:-Regression result of large firms
IJSER 2011
26

ISSN 2229-5518
CONCLUSION: in several dimensions. It verified the relationship for

Top 25 Indian pharmaceutical firms that are quoted
This paper extends earlier empirical studies
on the stock exchange of Mumbai for the year 1978 -
on the relationship between leverage and investment
IJSER 2011
27

ISSN 2229-5518
2009. Prior theoretical work posits that financial hack of funds. Hence we conclude that the leverage
leverage can have either a positive or a negative has no impact of pharmaceutical industry in India.
impact on the value of the firm because of its Cash flow and Retained earning play significant role
influence on corporate investment decisions. The in determing the investment the decisions due to the
investigation is motivated by the theoretical work of change in the monetary policy of the country. Cash
Myers(1977) Jen Seen (1986), Stulz (1988,1990) and flow effect investment decisions due to the
by an analytical work of McConnell and Servases imperfections of the capital market and due to the
(1990). We examined whether financing fact internal financing is cheaper than external
consideration affects firm investment decisions. We financing. These financing sources are far more
found that leverage is positively related to the level of important for small and highly leveraged firms. Our
investment and that this positive effect is results support Hite (1977) Who found that leverage
significantly stronger for firms with small firms and and investment are positively associated with given
negative impact on medium firms but positive impact the level of financing if an investment increase would
on large firms and this is not satirically significant. lower financial risk and hence the cost of bond
Further we inferred that the Indian pharmaceutical financing.
industry has heavy market demand for its product, so
that Industry had enormous cash flow and bloughing
REFERENCES
8. Johnson, Shane, A., (2003), Debt maturity and the effects of
1. Aivazian, V.A Callen, J.L., 1980, Corporate leverage and growth opportunities and liquidity risk on leverage, Review of
growth: the game theoretic issues, Journal of financial Financial Studies vol16, pp.209-236.
Economics 8, 379399.
9. Kopcke and Howrey, (1994), A panel study of investment:
2. Beush, T., Pagan, A., 1980,The language multiplier test and its Sales, cash flow, the cost of capital, and leverage, New England
applications to model specifications in econometirs, Review of Review, Jan/Feb., pp. 9-30
economic studies 47, 239, 253.Econommics of information and
Uncertainness, University of chicago press, chicago, pp 107-140. 10. Lang, L.E.; Ofek, E.; Stulz, R., (1996), Leverage, investment,
and firm growth, Journal of Financial Economics, Vol40, pp. 3-
3. Cantor; Richard, (1990), Effects of leverage on corporate 29.
investment and hiring decisions, Federal Bank of New York
Quarterly Review, pp. 31-41. 11. McConnell, John, J. and Servaes, H., (1995), Equity
ownership and the two faces of debt, Journal of Financial
4. Hausman, J.A., 1978, Specification tests in econometric, Economics, vol39, pp.131-157. .
Econometrica 46, 1251 1271.
12. Modigliani; Franco and Merton, H.; Miller (1958), The cost
5. Himmelberg, C.P., Hubbrad, R.G., Palia,D., 1999,
of Capital, corporation finance, and the theory of investment,
Understanding the determinants of managerial ownership and American Economic Review vol48, pp. 261-297.
the link between ownership and performance, Journal of financial
economics, 53, 353-384. 13. Modigliani; Franco and Merton, H.; Miller (1963), Corporate
income taxes and the cost of capital, a correction, American
6. Jensen; Michael, C., (1986), Agency costs of free cash flow, Economic Review vol48, pp. 261-297.
corporate finance and takeovers, American Economic Review
vol76, pp. 323-329. 14. Modigliani, F., Miller. M.H., 1958, The cost of capital,
corporation finance, and the theory of investment, American
7. Jensen, M.C., 1986, Ageny cost of free cash flow, corporate Economic Review 53, 433-443.
finance, and take-overs, American economic review 79, 323-329.
IJSER 2011
28

ISSN 2229-5518
15. Myers , S., 1977, Determinants of corporate borrowing,

Journal of financial Economics,5, 147-175.
16. Whited, T. (1992). Debt, Liquidity constraints and corporate

investment: Evidence from panel data, Journal of Finance vol 47,
pp.1425-1461.
Franklin John. S, The Principal, Nehru College of

Management, Coimbatore, Tamilnadu, India.
Email : [email protected]
Muthusamy.K, Research Scholar, School of
Management, Karunya University, Coimbatore,
Tamilnadu, India.
Email: [email protected]
IJSER 2011
29
ISSN 2229-5518
Effect of Nanofluid Concentration on the

Performance of Circular Heat Pipe
M. G. Mousa
Abstract The goal of this paper is to experimentally study the behavior of nanofluid to improve the performance of a circular heat
pipe. Pure water and Al2O3-water based nanofluid are used as working fluids. An experimental setup is designed and constructed to
study the heat pipe performance under different operating conditions. The effect of filling ratio, volume fraction of nano-particle in the
base fluid, and heat input rate on the thermal resistance is investigated. Total thermal resistance of the heat pipe for pure water and
Al2 O3-water based nanofluid is also predicted. An experimental correlation is obtained to predict the influence of Prandtl number and
dimensionless heat transfer rate, Kq on thermal resistance. Thermal resistance decreases with increasing Al2 O3-water based nanofluid
compared to that of pure water. The experimental data is compared to the available data from previous work. The agreem ent is found
to be fairly good.
Index Terms Heat Pipe, Thermal Performance, Nanofluids

Nomenclature
A Surface area, m2
FR Filling ratio
I Electric current, Amp
k thermal conductivity, W/m.K
N Number of thermocouples
Q Input heat rate, W
q Heat flux, W/m2
R Total thermal resistance of heat pipe, K/W
T Temperature, K
V Applied voltage, Volt
X Horizontal coordinate parallel to the test section, mm
Greek Symbols
Volume fraction of nanoparticles, %
Density, kg/m3
Dynamic viscosity, N.s/m2
Subscript
c Condenser
e Evaporator
ef Effective
e Evaporator
l Liquid
m Base fluid
n Nanofluid
p Particles
water Pure water
Dimensionless Numbers
RR Reduction factor in thermal resistance
K ef Le T
Kq Dimensionless heat transfer rate ,
Q
ef Cpef
Pr Prandtl number,
K ef
IJSER 2011
30
ISSN 2229-5518
INTRODUCTION
plate heat pipe (FPHP). Temperature fields in the heat pipe
T o solve the growing problem of heat generation by
were measured for different filling ratios, heat fluxes and
vapor space thicknesses. Experimental results showed that
electronic equipment, two-phase change devices such as heat the liquid distribution in the FPHP and consequently its
pipe and thermosyohon cooling systems are now used in thermal performance depends strongly on both the filling
electronic industry. Heat pipes are passive devices that ratio and the vapor space thickness. A small vapor space
transport heat from a heat source to a heat sink over thickness induces liquid retention and thus reduces the
relatively long distances via the latent heat of vaporization of thermal resistance of the system. Nevertheless, the vapor
a working fluid. The heat pipe generally consists of three space thickness influences the level of the meniscus
sections; evaporator, adiabatic section and condenser. In the curvature radii in the grooves and hence reduces the
evaporator, the working fluid evaporates as it absorbs an maximum capillary pressure. Thus, it must be, carefully,
amount of heat equivalent to the latent heat of vaporization. optimized to improve the performance of the FPHP. In all
The working fluid vapor condenses in the condenser and the cases, the optimum filling ratio obtained, was in the
then, returns back to the evaporator. Nanofluids, produced range of one to two times the total volume of the grooves. A
by suspending nano-particles with average sizes below 100 theoretical approach, in non-working conditions, was
nm in traditional heat transfer fluids such as water and developed to model the distribution of the liquid inside the
ethylene glycol provide new working fluids that can be used FPHP as a function of the filling ratio and the vapor space
in heat pipes. A very small amount of guest nano-particles, thickness.
when uniformly and suspended stably in host fluids, can Das et al. [7-8] and Lee et al. [9]) found great enhancement of
provide dramatic improvement in working fluid thermal thermal conductivity (5-60%) over the volume fraction range
properties. The goal of using nanofluids is to achieve the of 0.1 to 5%.
highest possible thermal properties using the smallest All these features indicate the potential of nanofluids in
possible volume fraction of the nano-particles (preferably < applications involving heat removal. Issues, concerning
1% and with particle size<50 nm) in the host fluid. stability of nanofluids, have to be addressed before they can
Kaya et al. [1] developed a numerical model to simulate the be put to use. Ironically, nanofluids of oxide particles are
transient performance characteristics of a loop heat pipe. more stable but less effective in enhancing thermal
Kang et al. [2], investigated experimentally, the performance conductivity in comparison with nanofluids of metal
of a conventional circular heat pipe provided with deep particles.
grooves using nanofluid. The nanofluid used in their study The aim of the present work is to investigate,
was aqueous solution of 35 nm diameter silver nano- experimentally, the thermal performance of a heat pipe. The
particles. It is reported that, the thermal resistance decreased affecting parameters on thermal performance of heat pipe are
by 10-80% compared to that of pure water. studied. The type of working fluid (pure water and Al2O3-
Pastukhov et al. [3], experimentally, investigated the water based nanofluid), filling ratio of the working fluid,
performance of a loop heat pipe in which the heat sink was volume fraction of nano-particles in the base fluid, and heat
an external air-cooled radiator. The study showed that the input rate are considered as experimental parameters.
use of additional active cooling in combination with loop Empirical correlation for heat pipe thermal performance,
heat pipe increases the value of dissipated heat up to 180 W taking into account the various operating parameters, is
and decreases the system thermal resistance down to 0.29 presented.
K/W. 2. Experimental Setup and Procedure
Chang et al. [4] investigated, experimentally, the thermal A schematic layout of the experimental test rig is
performance of a heat pipe cooling system with thermal shown in Fig.1. This research adopts pure water and Al2O3-
resistance model. An experimental investigation of water based nanofluid as working fluids. The size of nano-
thermosyphon thermal performance considering water and particles is 40 nm. The test nanofluid is obtained by
dielectric heat transfer liquids as the working fluids was dispersing the nano-particles in pure water. The working
performed by Jouhara et al. [5]. The copper thermosyphon fluid is charged through the charging line (6). In the heat
was 200 mm long with an inner diameter of 6 mm. Each pipe, heat is generated using an electric heater (12). The
thermosyphon was charged with 1.8 ml of working fluid and vapor generated in the evaporator section (8) is moved
tested with an evaporator length of 40 mm and a condenser towards the condenser section (4) via an adiabatic tube (5)
length of 60 mm. The thermal performance of the water whose diameter and length are 20 mm and 40 mm,
charged thermosyphon is compared with the three other respectively. Both evaporator and condenser sections have
working fluids (FC-84, FC-77 and FC-3283). The parameters the size of 40 mm-diameter and 60 mm-height. The
considered were the effective thermal resistance as well as condensate is allowed to return back to evaporator section by
the maximum heat transport. These fluids have the capillary action "wick structure" through the adiabatic tube.
advantage of being dielectric which may be better suited for The surfaces of the evaporator section, adiabatic section, and
sensitive electronics cooling applications. Furthermore, they condenser section sides are covered with 25 mm-thick glass
provide adequate thermal performance up to approximately wool insulation (3). Seventeen calibrated cooper-constantan
50 W, after which liquid entrainment compromises the thermocouples (T-type) are glued to the heat pipe surface
thermosyphon performance. and distributed along its length to measure the local
Lips et al. [6], studied experimentally, the performance of temperatures (Fig. 2). Two thermocouples are used to
IJSER 2011
31
ISSN 2229-5518
measure ambient temperature. All thermocouples are maintain the continuity of the interfacial evaporation,
connected to a digital temperature recorder via a multi-point capillary pressure must satisfy the following relation [2]
switch. The non condensable gases are evacuated by a
vacuum pump. The heat pipe is evacuated u to 0.01 bar via P Pl Pv Pe Pc
the vacuum line (10). The power supplied to the electric
(1)
heater (12) is measured by a multi-meter (13). The input
voltage was adjusted, using an autotransformer (2). The
Boiling limit is directly related to bubble formation in the
voltage drops across the heater were varied from 5 to 45
liquid. In order that a bubble can exist and grow in liquid, a
Volts. The A.C. voltage stabilizer (1) is used to ensure that
certain amount of superheat is required. Accurately
there is no voltage fluctuation during experiments. The
characterizing the thermal power transfer, Q is a complicated
pressure inside the evaporator was measured by a pressure
task because it is difficult to accurately quantify the energy
gage with a resolution of 0.01 bar.
loss to the ambient surroundings. Therefore, the whole
Thermocouples (with the uncertainty lower than 0.20 oC) are
surface of the heat pipe is well insulated so that the rate of
distributed along the surfaces of the heat pipe sections as
heat loss can be ignored. The heat input rate can be
follows: six thermocouples are attached to the evaporator
calculated using the supplied voltage and measured current
section, two thermocouples are attached to the adiabatic
such that,
section, and nine thermocouples are attached to the
condenser section. The obtained data for temperatures and
input heat rate are used to calculate the thermal resistance. Q I V
One can define the filling ratio, FR, as the volume (2)
of charged fluid to the total evaporator volume. The Where V and I are the applied voltage in Volt and current in
working fluid is charged at 30 oC. Amp, respectively.
The effects of working fluid type, filling ratio, volume
fraction of nano-particles in the base fluid, and heat input The experimental determination of the thermal
rate on the thermal performance of the heat pipe are performance of the heat pipe requires accurate
investigated in the experimental work. The experimental measurements of evaporator and condenser surface
runs are executed according to the following steps: temperatures as well as the power transferred. Calculating
1. The heat pipe is evacuated and charged with a certain evaporator and condenser temperatures is, relatively, a
amount of working fluid straightforward task. They are obtained by simply averaging
2. The supplied electrical power is adjusted manually at the the temperature measurements along the evaporator and
desired rate using the autotransformer. condenser surfaces. Thus, evaporator and condenser
3. The steady state condition is achieved after, temperatures can be expressed as:
approximately, one hour of running time using necessary i N te i N tc
adjustments to the input heat rate. After reaching the steady
state condition, the readings of thermocouples are recorded,
Ti 1
ei T
i 1
ci
sequentially, using the selector switch. The voltage of the

Te , Tc
Ne Nc
heater is measured to determine the value of applied heat
(3)
flux. Finally, the pressure inside the evaporator is recorded.
Where; Ne, Nc are the number of thermocouples on the
4. At the end of each run, power is changed and step 3 is
evaporator and condenser, respectively. The obtained data
repeated.
for temperatures and heat input rate are then used to
5. Steps 1 through 4 are repeated using with another adjusted
calculate the thermal resistance using the following relation,
amount of working fluid. The filling ratios, FR, used are 0.2,
0.4, 0.45, 0.50, 0.55, 0.60, 0.65, 0.70, 0.80 and 1.0. (Te Tc )
Pure water and Al2 O3-water based nanofluid are used as R
Q (4)
working fluids. Steps 1 through 5 are repeated for Al2O3-
water based nanofluid using several values of volume One can assume that the nano-particles are well
fractions of nano-particles. The volume fractions used are dispersed within the base-fluid, so the effective
0.25%, 0.4%, 0.5%, 0.6%, 0.75%, 1.0% and 1.5%, respectively. physical properties are described by classical
formulas which are mentioned by Das et al. [10] as;
3. Data Reduction
ef 1 m p
The effective dynamic viscosity of nanofluids can be
Although heat pipes are very efficient heat transfer
calculated using different existing equations that have been
devices, they are subject to a number of heat transfer
obtained for two-phase mixtures. The following relation is
limitation. For high heat flux heat pipes operating in low to
the well-known Einsteins equation for a viscous fluid
moderate temperature range, the capillary effect and boiling
containing a dilute suspension of small, rigid, spherical
limits are commonly the dominant factor. For a given
particles.
capillary wick structure and working fluid combination, the
pumping ability of the capillary structure to provide the ef m (1 2.5 )
circulation for a given working fluid is limited. In order to (6)
IJSER 2011
32
ISSN 2229-5518
k ef k m (1 7.4 ) using pure water.

The effect of adding nano-particles on the thermal
Where, is the ratio of the volume of the nano-particle to the
performance of the heat pipe is more evident if the data is
volume of the base fluid. expressed as a plot of the reduction rate in total thermal
The symbols Km and are the base fluid thermal resistance, RR versus , as shown in figure 7, i.e. the
conductivity and volume fraction of the nano-particle in the enhancement of thermal performance is increasing with the
base fluid respectively.
increase of nano-particles concentration, . The addition of
The relevant thermo-physical properties of the solid nano-
nano-particles has illustrated that during nucleate boiling
particles (Al2O3) used in the present study are Cpp =773
some nano-particles deposit on the heated surface to form a
J/kg. oC, =3880 kg/m3, and kp =36 W/m. oC, which are
porous layer. This layer improves the wet ability of the
mentioned in previous work [10].
surface considerably. The thermal conductivity of the
working fluid is also preferably high in order to minimize
One can calculate the reduction factor in total
the temperature gradient.
thermal resistance of heat pipe charged with nanofluid by
The obtained heat transfer data is correlated as the
referring its thermal resistance to that charged with pure
following relation:
water, expressed as;
RR= (Rwater Rn)/Rwater (7)
R 0.294 K q0.596 FR 1.273 Pr 0.0532 (8)
The maximum relative error in total thermal The error in calculated thermal resistance is predicted by the
resistance is about + 11.9 %. above suggested correlation is around 5 %, as shown in Fig.
8.
4. Result and Discussion 4.1. Comparison with the available literature
Experiments are performed on heat pipe considering two Figure 9 shows a comparison between the present
different working fluids; pure water and Al2O3-water based experimental results with those reported by Kang [2] in case
nanofluid. In both cases, the effect of heat input rate, Q , and of using pure water as a working fluid with FR equals to 0.5.
filling ratio, FR on its performance are investigated. It can be observed that the present experimental results for
Moreover, in case of the nanofluid, the effect of varying the used two working fluids have the same trend as those
volume fraction of the nano-particles in the base fluid, , on reported by Kang [2]. The difference between both results
the thermal performance of this heat pipe is also predicted. when using pure water may be attributed to the difference in
The values of local surface temperatures along all sections of heat pipe geometry and uncertainty in measurements.
the heat pipe are measured (0 X 120 mm), where X is Figure10 shows a comparison between the present
measured from the beginning of the evaporator section. The experimental results of the total resistance, R using Al2O3-
range 0.0 X 40 mm represents the evaporator section, 40 < water based nanofluid of =0.5% to those reported by Lips
X 80 mm represents the adiabatic section and 80<X120 and Lefvre [6] who used a heat pipe charged with n-
mm represents the condenser section. pentane nanofluid of = 0.5%. One can see that the thermal
Figure 3 and 4 illustrate the surface temperature along the resistance of the heat pipe decreases with increasing filling
heat pipe for two values of heat input rate (40 and 60 W) and ratio up to a value of FR = 0.45, where it starts to increase
same filling ratio, FR of 0.45. As expected, the surface with increasing the filling ratio. It can be also observed that
temperature decreases with increasing the the present experimental results are, a little bit, higher than
those reported by Lips et al [6], but they have the same
distance from the evaporator section due to the heat trend. The discrepancies in both results may be due to the
exchange between the heat pipe surface and working fluid. differences in dimensions of the tested heat pipes as well as
It is clear that the surface temperature decreases with due to the type of working fluids used.
increase of nano-particle concentration, . As expected, the 5. Conclusions
surface temperature decreases with increasing the distance Using heat pipes and based on the nanofluid literature,
from the evaporator along heat pipe. particularly those related to the optimum operating
Figure 5 illustrates the variation of the total thermal condition, the thermal performance enhancement of heat
resistance of the heat pipe, R with the filling ratio for Al2O3- pipes charged with the nanofluids indicates the potential of
water based nanofluid =0, 0.5% and 1.2 %) at heat input the nanofluid use as substitute of conventional fluids. This
rate of 40 W. As shown in the figure, R decreases with the finding makes the nanofluid more attractive as a cooling
increase of the filling ratio up to a value of FR equals to 0.40, fluid for devices with high power intensity. A compact heat
after which R starts to increase with the increase of FR due to pipe is thermally tested with two different working fluids;
increasing liquid inside evaporator. It can be also noticed pure water and Al2O3-water based the nanofluid. The
that the thermal resistance, R is inversely proportional to thermal performance of this heat pipe is predicted under
concentration of nano- particles, . different operating conditions including heat input rate,
Figure 6 shows the variation of the total thermal filling ratio, and volume fraction of the nano-particle in
resistance of the heat pipe, R with volume fraction of nano- water. From the obtained data and its discussion, the
particle in the base fluid, at two different heat input rates, following conclusions may be drawn:
1- The optimum filling ratio of charged fluid in the tested heat
Q (40 and 60 W). Over the tested range of , while keeping
pipe was about 0.45 to 0.50 for both pure water and
the filling ratio of 0.45, the percentage enhancement in R
reaches up to 62.60% at compared to its value when
IJSER 2011
33
ISSN 2229-5518
Al2O3-water based the nanofluid, respectively.

2- By increasing concentration of the nanofluid, the thermal
performance of heat pipe can be decreased
It can be said that better ability to manage thermal
properties of working fluid translates into greater energy
transport, smaller and lighter thermal systems. This may be
applied to cooling of super computers.
References
[1] T. Kaya, R. Perez, C. Gregori and A.Torres, "Numerical
simulation of transient operation of loop heat pipes", Applied
Thermal Engineering, vol.28, 2008, pp.967-974.
[2] S. W. Kang, W. C. Wei, S. H. Tsai, and S. Y. Yang,
"Experimental investigation of silver nanofluid on heat pipe
thermal performance", Applied Thermal Engineering, vol.26,
2006, pp.23772382.
[3] V.G. Pastukhov, Y. F. Maidanik, C.V. Vershinin, and
M.A. Korukov, "Miniature loop heat pipes for electronics
cooling", Applied Thermal Engineering, vol.23, 2000,
pp.1125-1135.
[4] Y. W. Chang, C. H. Cheng, J. C. Wang, and S. L. Chen,"
Heat pipe for cooling of electronic equipment" Energy
Conversion and Management, 2008.
[5]H. Jouhara1, O. Martinet, and A.J.
Robinson, Experimental Study of Small Diameter
Thermosyphons Charged with Water", FC-84, FC-77 & FC-
3283- 5th European Thermal-Sciences Conference, The
Netherlands, 2008.
[6] S. Lips, F. Lefvre, and J. Bonjou, "Combined effects of
the filling ratio and the vapour space thickness on the
performance of a flat plate heat pipe", International Journal
of Heat and Mass Transfer vol. 53, 2010, pp. 694702.
[7] S. K. Das, U.S. Choi, W. Yu, and T. Pradeep, "Nanofluid
Science and Technology", Wily-Interscience, 2007.
[8] Das, S. K., Putra, N., Thiesen, P., and Roetzel, W.,
"Temperature dependence of thermal conductivity
enhancement for nanofluids", J. Heat Transfer vol.125, 2003,
pp. 567-574.
[9] R.B. Mansour, N. Galanis, and C.T. Nguyen, " Effect of
uncertainties in physical properties on forced convection heat
transfer with nanofluids," Applied Thermal Engineering Vol.
27, 2007, pp. 240249.
[10] S. Lee, S.U.S Choi, S. Li., and J.A. Eastman, "Measuring
thermal conductivity of fluids containing oxide
nanoparticles," J. Heat Transfer vol. 121, 1999, pp. 280-289.
IJSER 2011
34

ISSN 2229-5518
1. Stabilizer 2. Autotransformer 3. Insulation 4.condenser 5.Adiabatic tube

6. Charge line 7. Pressure gage 8. Evaporator 10. Vacuum line
11. Mica sheet 12. Electric heater 13. Multi-meter
Figure 1 Schematic layout of the test rig
Figure 2 Thermocouples distribution along the heat pipe sections
IJSER 2011
35
INTERNATIONAL JOURNAL OF SCIENTIFIC & ENGINEERING RESEARCH, VOLUME , ISSUE, - 7
ISSN 2229-5518
42 0.2
F.R = 0.45, Q=40 W FR=0.40
Thermal resistance, R, K/W

Q = 40 W
Q = 60 W

40
0.16

Temperature, C
o
38
0.12
36
0.08
34
0.04
0 0.4 0.8 1.2
32
0 40 80 120
Volume Fractionof nanoparticle,
Distance, mm
Figure 6 Variation of thermal resistance with volume
Figure 3 Temperature distribution along the heat fraction of nano-particles.
pipe surface for different nanofluid concentration 80
46 Q=40 W
(R water - R nanofluid )/R water, %

Q=60 W
F.R = 0.45, Q=60 W

44
60

Temperature, C
42
o
40
40
38
20
36
0
34 0 0.4 0.8 1.2
0 40 80 120
Distance, mm Volume fraction ofNano Particles,

Figure 4 Temperature distribution along the heat Figure 7 Reduction factor of total thermal resistance at
pipe surface for different nanofluid concentration different volume fraction of nano-particles,
0.24
0.28
Experimental Thermal Resistance, K/W
Q=40 W

0.2
0.24
0.16
0.2
0.12
0.16 0.08
0.04
0.12
0.2 0.4 0.6 0.8 1
F i l l i n g R a t i o, FR 0
0 0.05 0.1 0.15 0.2 0.25
Figure 5 Variation of thermal resistance with
filling ratio at different heat input rates Corelated Thermal Resistance, K/W
Figure 8 Experimental Nusselt number versus correlated
IJSER 2011
36
INTERNATIONAL JOURNAL OF SCIENTIFIC & ENGINEERING RESEARCH, VOLUME , ISSUE, - 8
ISSN 2229-5518
Nusselt number over the tested range of heat input rate

0.24
0.2
Thermal Resistance ( oC / W)
0.16
0.12
0.08 Present Study (

Present Study (
Kang et al [2 ](
0.04
0 20 40 60
Heat Input ( W )
Figure 9 Comparison of the present results with

available literature at different value of
heat input rate
0.36
Q=10 W
Lips et al.[6],=0.5% n-petane
0.32 Present Study =0.5%, Al2O3
0.28
0.24
0.2
0.16
0.12
0 0.2 0.4 0.6 0.8 1
F i l l i n g R a t i o, FR
Figure 10 Comparison between present data and
previous one for the variation of thermal
resistance with filling ratio
IJSER 2011
37
ISSN 2229-5518
Enhancement of Person Identification using Iris

Pattern
Vanaja roselin.E.Chirchi, Dr.L.M.Waghmare , E.R.Chirchi
Abstract The biometric person identification technique based on the pattern of the human iris is well suited to be applied to access
control. Security systems having realized the value of biometrics for two basic purposes: to verify or identify users. In this busy world,
identification should be fast and efficient. In this paper we focus on an efficient methodology for identification and verification for iris
detection using Haar wavelet and the classifier used is Minimum hamming distance, even when the images have obstructions, visual noise
and different levels of illuminations.
Index TermsBiometrics, Iris identification, Haar wavelet, occluded images, veriEye.

1. Introduction the criminals or malicious intruders to fool recognition
system or program. Iris cannot be spoofed easily.
B iometrics which refers to identifying an individual

by his or her physiological or behavioral Palm Print: Palm print verification is a slightly
modified form of fingerprint technology. Palm print
characteristics has capability to distinguish between
authorized user and an imposter. An advantage of using scanning uses an optical reader very similar to that
biometric authentication is that it cannot be lost or used for fingerprint scanning; however, its size is much
forgotten, as the person has to be physically present bigger, which is a limiting factor for use in
during at the point of identification process workstations or mobile devices.
[9].Biometrics is inherently more reliable and capable Signature Verification: It is an automated method of
than traditional knowledge based and token based examining an individuals signature. This technology is
techniques. The commonly used biometric features dynamic such as speed, direction and pressure of
include speech, fingerprint, face, Iris, voice, hand writing, the time that the stylus is in and out of contact
geometry, retinal identification, and body odor with the paper. Signature verification templates are
identification [10] as in Fig.1 typically 50 to 300 bytes. Disadvantages include
problems with long-term reliability, lack of accuracy and
cost.
Fingerprint: A fingerprint as in Fig.1 recognition
system constitutes of fingerprint acquiring device,
minutia extractor and minutia matcher. As it is more
Fig. 1: Examples of Biometrics common biometric recognition used in banking, military
etc., but it has a maximum limitation that it can be
To choose the right biometric to be highly fit for the
spoofed easily. Other limitations are caused by
particular situation, one has to navigate through some
particular usage factors such as wearing gloves, using
complex vendor products and keep an eye on future
cleaning fluids and general user difficulty in scanning.
developments in technology and standards. Here comes
Iris Scan: Iris as shown in Fig.2 is a biometric
a list of Biometrics with comparatives:
feature, found to be reliable and accurate for
Facial Recognition: Facial recognition records the authentication process comparative to other biometric
spatial geometry of distinguishing features of the face. feature available today. As a result, the iris patterns in
Different vendors use different methods of facial the left and right eyes are different, and so are the iris
recognition, however, all focus on measures of key patterns of identical twins. Iris templates are typically
features of the face. Facial recognition has been used in around 256 bytes. Iris scanning can be used quickly for
projects to identify card counters or other undesirables both identification and verification applications because
in casinos, shoplifters in stores, criminals and terrorists of its large number of degrees of freedom. Iris as in Fig. 2
in urban areas. This biometric system can easily spoof by is like a diaphragm between the pupil and the sclera and
IJSER 2011
38

ISSN 2229-5518
its function is to control the amount of light entering further implementation we will be using Chinese
through the pupil. Iris is composed of elastic connective academy of science-Institute of automation (CASIA) iris
tissue such as trabecular meshwork. The agglomeration image database available in the public domain [7].
of pigment is formed during the first year of life, and 3. Preprocessing
pigmentation of the stroma occurs in the first few years 3.1 Locating Iris
[7][8]. The first processing step consists in locating the inner
and outer boundaries of the iris and second step to
normalize iris and third step to enhance the original
image as in Fig.4 [4][6]. The Daugmans system, Integro
differential operators as in (1) is used to detect the center
and diameter of iris and pupil respectively.
2
max(r, x0, y0)
r 0 I (r * cos x0, r * sin y 0) --- (1)

Fig. 2: Structure of Eye Where (x0, y0) denotes the potential center of the
searched circular boundary, and r its radius.
The highly randomized appearance of the iris makes its 3.2 Cartesian to polar reference transform
use as a biometric well recognized. Its suitability as an Cartesian to polar reference transform suggested by
exceptionally accurate biometric derives from [4]: J.Daugman authorizes equivalent rectangular
i. The difficulty of forging and using as an
representation of the zone of interest as in Fig.4, remaps
imposter person;
each pixel in the pair of polar co-ordinates(r, ) where r
ii. It is intrinsic isolation and protection from the and are on interval [0,1] and [0,] respectively. The
external environment;
unwrapping in formulated as in (2) [1].
iii. Its extremely data-rich physical structure;
iv. Its genetic propertiesno two eyes are the I ( x(r, ), y(r, )) I (r , ) --- (2)
same. The characteristic that is dependent on
Such that
genetics is the pigmentation of the iris, which
determines its color and determines the gross x(r, ) (1 r ) xp( ) rxi( ),
anatomy. Details of development, that are --- (3)
y(r , ) (1 r ) yp( ) ryi( )
unique to each case, determine the detailed
morphology; where I(x, y), (x, y), (r, ), (xp, yp), (xi, yi) are the
v. its stability over time; the impossibility of iris region, Cartesian coordinates, corresponding polar
surgically modifying it without unacceptable coordinates, coordinates of the pupil, and iris
risk to vision and its physiological response to boundaries along the direction, respectively.
light, which provides a natural test against
Image Acquisition
artifice.
After the discovery of iris, John G. Daugman, a
professor of Cambridge University [8] ,[9], suggested an Preprocessing
image-processing algorithm that can encode the iris
pattern into 256 bytes based on the Gabor transform.
In general, the iris recognition system is composed Feature Extraction
of the following five steps as depicted in Fig. 3
According to this flow chart, preprocessing including
image enhancement. Data Base : Reference Code Pattern Matching
2. Image Acquisition. Identify or Reject Subject

An image of the eye to be analyzed must be
acquired first in digital form suitable for analysis. In
IJSER 2011
39

ISSN 2229-5518
Fig. 3: General steps of the iris recognition system process will be fast and simple accordingly. Performance
of classifiers is based on minimum Hamming Distance
(MHD) as in (4).
HD=XOR (codeA, codeB) --- (4)
Where codeA and codeB are the templates of two

(a) (b) images. When an iris image is captured in system, the
designed classifier compares it with the whole images in
each class. The Hamming distance (HDs) between input
(c) images and images in each class are calculated, then the
two different classifiers are being applied as follows
[1][4].
I. In the first classifier, the minimum HD between
(d) input iris code and codes of each class is
Fig.4: (a) Original image; (b) localized iris; (c) normalized iris; computed.
And (d) enhanced iris.
II. In the second classifier, the harmonic mean of
the n HDs that have been recorded yet is
assigned to the class as in (5)[4].
4. Feature Extraction
The most important step in automatic iris length(code)
recognition is the ability of extracting some unique HM --- (5)
i 1 (code) (1 / code(i))
length
attributes from iris, which help to generate a specific
code for each individual. Gabor and wavelet transforms
are typically used for analyzing the human iris patterns 6. Identification and Verification
and extracting features from them, In Fig. 5(a) [1],[2] ,[4] Identification and verification modes are two main
a conceptual chart of basic decomposition steps for an goals of every security system based on the needs of the
image is depicted. The approximation coefficients matrix environment. In the verification stage, the system checks
cA and details coefficients matrices cH, cV and cD
if the user data that was entered is correct or not (e.g.,
(horizontal, vertical, and diagonal, resp.) Obtained by
username and password) but in the identification stage,
wavelet decomposition of the input image are shown in
Fig. 5(b) [4],[11],[12]. The definitions used in the chart the system tries to discover who the subject is without
are as follows. any input information. Hence, verification is a one-to-
(i) C denote downsample columnskeep the even one search but identification is a one-to-many
indexed columns. comparison.
(ii) D denote downsample rowskeep the even-
indexed rows. 7. Application
(iii) Lowpass D denotes the decomposition lowpass
Implementations of image processing above five
filter.
steps as in Fig.3 is used in veriEye iris recognition
(iv) Highpass D denotes the decomposition highpass
technology for healthcare patient identification system
filter.
in 2008[5].Algorithm advanced iris segmentation,
(v) The blocks under Rows convolve with filter of
enrollment and matching which is as follows [5].
block the rows of entry.
Robust eye iris detection: Irises are detected even when
(vi) The blocks under Columns convolve with filter of
the images have obstructions, visual noise and different
block the columns of entry.
levels of illumination. Lighting reflections, eyelids and
(vii) Ii denotes the input image.
eyelashes obstructions are eliminated. Images with
5. Pattern Matching narrowed eyelids or eyes that are gazing away are also
By comparing similarities between feature vectors of accepted using wavelet algorithm[5].
two irises to determine that they are accepted or
rejected. Since feature vector is binary, the matching
IJSER 2011
40

ISSN 2229-5518
Automatic interlacing detection and correction: The Size of one record in a

2.3 Kbytes
correction results in maximum quality of iris features database
templates from moving iris images.
Gazing-away eyes: A gazing-away iris image is Maximum database size Unlimited
correctly detected, segmented and transformed as if it
were looking directly into the camera.
Correct iris segmentation: is achieved under these
conditions: 8. CONCLUSION
i. Perfect circles fail. VeriEye uses active shape The research work is to enhance the algorithm for
models that more precisely model the contours efficient person identification for other area of
of the eye, as perfect circles do not model iris
applications by increasing FRR more than 0.33% as the
boundaries.
ii. The centers of the iris inner and outer VeriEye algorithm results with FRR 0.32% and FAR
boundaries are different Fig. 8. The iris inner 0.001%. Wavelets iris recognition algorithm is suitable
boundary and its center are marked in red; the for reliable, fast and secure person identification.
iris outer boundary and its center are marked in Wavelet, Gabor filter and hamming distance are used in
green. veriEye algorithm for robust and fast matching for
iii. Iris boundaries are definitely not circles and
healthcare application for patient identification.
even not ellipses Fig. 9.and especially in gazing-
Research work will also focus on the algorithm for rapid
away iris images.
iv. Iris boundaries seem to be perfect circles. The and accurate iris identification even if the images are
recognition quality can still be improved if occlude further algorithm will also focus on robust iris
boundaries are found more precisely Fig.10. recognition, even with gazing-away eyes or narrowed
Compared to perfect circular white contours eyelids which solves all the security related problems.
v. Fast matching. Configurable matching speed
varies from 50,000 to 150,000 comparisons
Reference:
per second. The highest speed still preserves
[1] Christel-loc TISSE1, Lionel MARTIN1, Lionel
nearly the same recognition quality Fig. 11.
TORRES 2, Michel ROBERT Person identification
vi. Reliability. VeriEye 2.0 algorithm shows
technique using human iris recognition.
excellent performance when tested on all
[2] D.E.Benn, M.S.Nixon and J.N.Carter, Robust eye
publicly available datasets Extraction using H.T. AVBPA99
[3] D.Gabor, theory of communication, journal IEEE.
7.1 Technical Specification[5] [4] A.Poursaberi and B.N.Araabi, iris recognition for
Minimal radius of circle partially occluded images: methodology and
64 pixels sensitivity Analysis Hindawi Publishing
containing full iris texture
Corporation EURASIP Journal on Advances in
Signal Processing Volume 2007, Article ID 36751, 12
Iris rotation tolerance 15 degrees
pages
[5] http:// www.neurotechnology.com
Recommended iris image [6] John Daugman. Recognizing persons by their iris
Near-infrared
capture spectral region patterns Cambridge University, Cambridge, UK.
[7] L.M. Waghmare, S. P. Narote, A.S. Narote,
Iris template extraction Biometric Personal Identification Using IRIS,
0.5 sec Proceedings of International Conference on
time
Systemics, Cybernetics and Informatics, ICSCI -
2006, Pentagram Research Centre Hyderabad, pp.
50,000 -
679-682, Jan 2006
Matching speed 150,000
[8] E. Wolff, Anatomy of the eye and orbit, H. K. Lewis,
irises/sec
London, UK, 7th edition, 1976.
IJSER 2011
41

ISSN 2229-5518
[9] John Daugman,How Iris works IEEE Transaction

on circuit and systems for Video Technology,
VOL.14, No.1, January 2004.
[10] D.Zang Automated biometrics technologies and
systems Klumer Academics, Boston, Mass, USA,
2000.
[11] A.S. Narote, S.P. narote, M.B. Kokare, L.M.
Waghmare, An Iris Recognition Based on Dual Tree
Complex Wavelet Transform published in the
proceedings of IEEE International Conference
TENCON 2007, Taiwan during Oct. 30-Nov. 02,
2007.
[12] R.W.Conners and C.A.Harlow, A theoretical
comparison of texture algorithms, IEEE transactions
on Pattern Analysis and Machine Intelligence, vol2,no.
3,pp.204-222, 1980.
IJSER 2011
42

ISSN 2229-5518
(a)
(b)
Fig.5 :( a) wavelet decomposition steps diagram and (b) 4-level decomposition of typical image with db2 wavelet
Fig.8 Fig.9 Fig.10 Fig.11
IJSER 2011
43

ISSN 2229-5518
Mn DOPED SnO2 Semiconducting Magnetic

Thin Films Prepared by Spray Pyrolysis Method
K.Vadivel, V.Arivazhagan, S.Rajesh
Abstract -- Semiconducting magnetic thin films of SnO2 doped with Mn was prepared by spray pyrolysis method. The polycrystalline nature of the
films with tetragonal structure was observed from X-ray Diffractometer. The calculated crystalline size was 16-22 nm and the lattice constant is
a=4.73A and c=3.17A. The compositional studies give the weight percentage of the used materials. The absorption edge starts with 294 nm and
rise in transmittance spectra shows the nanocrystalline effect of as deposited films. The calculated band gap from the absorption coefficient is 3.25
eV which greater than the bulk band gap of Tin oxide. The electrical properties of the prepared films also reported in this paper.
Index Terms -- Mn doped SnO2, Spray Pyrolysis, XRD, UV, Electrical study.
1. INTRODUCTION
THE study of SnO2 transparent conducting oxide thin films are

of great interest due to its unique attractive properties like high
impurities is expected to play an important role in changing the
charge carriers concentration of the metal oxide matrix, catalytic
activity, the surface potential, the phase composition, the size of
optical transmittance, uniformity, nontoxicity, good electrical, low
crystallites, and so on [8- 10]. It is expected that various
resistivity, chemical inertness, stability to heat treatment,
concentration of Mn in SnO 2 may affect the structural, optical and
mechanical hardness, Piezoelectric behavior and its low cost. SnO2
magnetic properties of the films. From bang gap engineering point
thin films have vast applications as window layers, heat reflectors in
of view, suitable band gap is essential for the fabrication of optical
solar cells, flat panel display, electro-chromic devices, LEDS, liquid
devices. So far our knowledge is concerned there are very few
crystal displays, invisible security circuits, various gas sensors etc.
reports available on the deposition of Mn doped SnO2 thin films by
Undoped and Cu, Fe and Mn doped SnO2 thin films have been
spray pyrolysis method. In considering the importance of these
prepared by vapor deposition technique and reported that SnO 2
materials in the field of magnetic materials, we have prepared Mn
belongs to n-type semiconductor with a direct optical band gap of
doped SnO 2 films using a simple and locally fabricated spray
about 4.08 eV [6]. To improve the quality of the films as well as the
pyrolysis system relatively at the temperature of 450C.
physical and chemical properties, the addition of some metal ions as
2. EXPERIMANTAL
Mn doped SnO2 thin films were prepared by spray for preparing nanocrystalline films plays an important role. Here the
pyrolysis method. Mn doped SnO2 thin films were prepared by temperature of the substrate kept at 450C and the solution was
spray pyrolysis method. The starting materials were SnCl4.5H2O sprayed using atmospheric air as carrier gas. Then the film was
for Tin and Mn(CHOO3)2.4H2O for Manganese.The allowed to natural cool down. The structural studies on as deposited
concentration of 0.5m of Stannous chloride and 0.1m of Manganese manganese doped tin oxide thin films were analyzed using X-Ray
acetate was taken in two different beakers with double distilled diffractometer (Shimadzu XRD-6000). Using EDAX (JSM
water. Then 98% of Stannous chloride solution and 2% of 6390) the elemental composition of the films was carried out. The
manganese acetate solution was mixed together and stirred using optical and electrical properties of the films done by UV-Vis
magnetic stirrer for 4 hours and allowed to aging for ten days. The spectrometer (Jasco-570 UV/VIS/ NIR) and Hall (Ecopia HMS-
clear solution of the mixer was taken for film preparation by spray 3000) measurement system.
pyrolysis method. The temperature of the substrate in this method
IJSER 2011
44

ISSN 2229-5518
3. RESULT AND DISCISSION
3.1. Structural studies on Mn:SnO2 thin films
The structural studies on as deposited Mn doped Tin from large number of diffracted peaks. The tetragonal structure of
oxide were analyzed by X-Ray spectrophotometer and the graph the sample with the three strong peaks of (1 1 0), (1 0 1) and (2 1 1)
between 2 theta versus diffracted ray intensity is shown in figure 1. correspond with peak position of 2=26.3609, 33.6541 and 51.6145
The polycrystalline natures of the prepared samples were observed respectively were identified using standard JCPDS files.
Fig 1. XRD pattern of Mn doped SnO2 nanocrytalline thin films
Substrate temperature is one of the main parameters, which of as deposited films were calculated using Debye-Scherrers
determine the structural properties of the films. The crystalline size formula given by,
--------------------- (1)
Where is the wavelength of X-ray used (1.54 A), is the full angle. The lattice constant of the spray coated Tin oxide
width half maximum (FWHM) of the peak and is the glancing films calculated using the formula
,
-------------------- (2)
Whered is the interplanar distance, (h k l) are the Miller indices The calculated crystalline size (D) and lattice constant (a and c) of
and a and c are the lattice constant for the Tetragonal structure. spray coated Mn doped Tin oxide are tabulated in table 1.
Table.1. Structural parameters of Mn doped SnO2
Substrate hkl d 2 FWHM D c a

temperature (A ) () (A ) (A ) (A )
110 3.3420 26.6517 0.3699 220
450 C 101 2.6342 34.0089 0.4690 177 3.17 4.73
211 1.7608 51.8835 0.5271 167
IJSER 2011
45

ISSN 2229-5518
Fig 2. The EDAX spectrum of as deposited Mn doped Tin Oxide thin films
Fig 2 shows,The EDAX spectrum shows the compositional wt % of atomic percentage of doped Mn was observed as 1.09 % and 0.45 %
the used materials. The weight and atomic percentage of Sn was respectively.
observed as 122.96 % and 23.65 % respectively. The weight and
3.3. Optical studies on SnO2 thin film
The optical studies of the Mn doped films were studies by the films. Also the absorption peaks around 400 nm and 550 nm
UV Vis spectrometer in the range of 200-900 nm. The absorption (indicated by arrow) observed in the graph shown in fig 3.
edge starts with 294 nm reveals that the Nanocrystalline effect of
1.6
45
1.4 40
35
1.2
30
Transmittance %
Absorption %
1.0
25
0.8 20
15
0.6
10
0.4
5
0.2 0
300 400 500 600 700 800 300 400 500 600 700 800 900
Wavelength (nm) Wavelength (nm)
Fig 3. Absorption and transmittance spectra of Mn doped Tin oxide thin films
nanocrystalline effect of as deposited films. The optical band gap

of the film was calculated using absorption coefficient. The band
The light transmitted spectra have the maximum transmittance of 40 gap value could be obtained from the optical absorption spectra by
percentage is shown in fig 3. The rise in transmittance spectra using Taucs relation [12],
observed at around 500 nm and 700 nm may be due to
--------------------------- (3)
Where is the absorption coefficient, ( ) is the photon energy

and A is a constant. The direct band gap semiconductor
can be obtained from the relation,
IJSER 2011
46

ISSN 2229-5518
(4)
100
90
80
70
60
13
50
h) *10
2
40
30
20
10
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Photon energy (eV)
Fig 4. Optical band gap plot between photon energy versus absorption coefficient
Fig.4 shows the variation of ( 2

versus ( ) for the SnO2 observed in films prepared at 450C of substrate temperature. It
thin film. The straight nature of the films over the wide range of clearly shows the observed value of band gap is greater than the
photon energy indicates the direct type of transition. The optical gap bulk band gap (2.5 eV) of Tin oxide.
has then been determined by extrapolation of the linear region on

the energy axis shown in fig.4. The optical band gap of 3.25 eV was
3.3. Electrical properties of Mn doped SnO2 films
The electrical properties of the prepared films carried out as 2.161x103 -1 cm-1 and 4.628 x10-4 .cm respectively. The
using Hall measurement system at room temperature with the given carrier concentration of the Mn doped SnO2 have the value of
input voltage of 1 mA. The negative sign of the Hall coefficient minus 1.703x1021cm-2. The mobility of the films were found as
-3
value of -3.666x10 shows the n-type semiconducting nature of the 7.922 cm2/V.sec . From these result is observed that the Mn doped
films. The conductivity () and resistivity () of the film observed SnO2 films have good electrical properties.
4. CONCLUSION
Manganese doped Tin oxide thin films were prepared by The optical studies reveals that the presence of nanoparticle on the
spray pyrolysis method. The X-ray diffractogram shows the films. The signature of nanocrystalline effect of as deposited film is
polycrystalline nature of as deposited films with tetragonal absorption edge (294 nm) and the rise in transmittance spectra. The
structure. The crystalline size of the film was calculated using calculated band gap of 3.25 is greater than the bulk band of value of
Debye-Scherer formula is varies from 16-22 nm corresponds to Tin oxide. The n-type semiconducting nature of the films observed
three strong peaks. The calculated lattice constant of the films from from negative sign of the Hall coefficient. The conductivity of

interplanar distance and peak plane is a=4.73A and c=3.17A . 2.161x103 -1 cm-1 was observed on as deposited films.
REFERNCES
[1].Arivazhagan.V , Rajesh.S, Journal of Ovonic research, [3] R. S. Rusu, G. I. Russia, J. Optoelectron. Adv. Mater 7(2), 823
Vol.6,No.5 ,221-226 ,(2010) (2005).
[2] J. B. Yoo, A. L. Fahrenbruch, R. H. Bube, J Appl Phys. 68, 4694 [4] M. Penza, S. Cozzi, M. A. Tagliente, A. Quirini, Thin Solid
(1990). Films, 71, 349 (1999).
IJSER 2011
47

ISSN 2229-5518
[5] S. Ishibashi, Y. Higuchi, K. Nakamura, J. Vac. Sci. Technol., [13]. Badawy W.A et al Electrochem, Soc., Vol.137, 1592-
A8, 1403 (1998). 1595,(1990)
[6] J. Joseph, V, K. E. Abraham, Chinese Journal of Physics, 45, [14]. Bruneaux J et al,Thin Solid Films, Vol.197, 129-142,(1991)
No.1, 84 (2007). [15]. Chitra Agashe et al. J. Appl. Phys. Vol.70, 7382-7386,(1991)
[7] E. Elangovan, K. Ramamurthi, Cryst. Res. Technol., 38(9), 779 [16]. Chitra Agashe et al, Solar Energy Mat.,Vol.17 ,99-117,(1988)
(2003). [17]. Datazoglov O. Thin Solid Films, Vol.302, 204-213,(1997)
[8]. Datazoglov O. Thin Solid Films, Vol.302, 204-213,(1997) [18]. Fantini M. and Torriani I. Thin Solid Films, Vol.138, 255-265
[9]. Fantini M. and Torriani I. Thin Solid Films, Vol.138, 255-265 ,(1986).
,(1986). [19]. Garcia F.J., Muci J. and Tomar M.S. Thin Solid Films, Vol.97,
[10]. Garcia F.J., Muci J. and Tomar M.S. Thin Solid Films, Vol.97, 47-51,(1982)
47-51,(1982) [20]. Ghoshtagore R.N. J. Electrochem. Soc., Vol.125, 110-
[11]. Z. C. Jin, J. Hamberg, C. G. Granqvist, J Appl Phys.64, 5117 17,(1978)
(1988). [21]. Segal and Woodhead J L Proc.Br.Ceram.Soc.38, 245, 1986
[12].Advani G.N et al, Thin Solid Films, 361 367,(1974)
*Corresponding Author
K.Vadivel*, V.Arivazhagan, S.Rajesh- Research Department of
Physics, Karunya University, Coimbatore, Tamilnadu, India-641
114. *Email: [email protected]
IJSER 2011
48
ISSN 2229-5518
A Novel comprehensive method for real time

Video Motion Detection Surveillance
Sumita Mishra, Prabhat Mishra, Naresh K Chaudhary, Pallavi Asthana
Abstract This article describes a comprehensive system for surveillance and monitoring applications. The development of an
efficient real time video motion detection system is motivated by their potential for deployment in the areas where security is the main
concern. The paper presents a platform for real time video motion detection and subsequent generation of an alarm condition as set by
the parameters of the control system. The prototype consists of a mobile platform mounted with RF camera which provides continuous
feedback of the environment. The received visual information is then analyzed by user for appropriate control action, thus enabling the
user to operate the system from a remote location. The system is also equipped with the ability to process the image of an object and
generate control signals which are automatically transmitted to the mobile platform to track the object.
Index Terms Graphic User Interface, object tracking, Monitoring, Spying, Surveillance, video motion detection
1 INTRODUCTION
V
ideo Motion Detection Security Systems (VMDss) making use of the difference of consecutive frames
have been available for many years. Motion (two or three) in a video sequence.
detection is a feature that allows the camera to This method is highly adaptive to dynamic
detect any movement in front of it and transmit the environments hence it is suitable for present
image of the detected motion application with certain modification. Presently
to the user. VMDss are based on the ability to respond advanced surveillance systems are available in the
to the temporal and/or spatial variations in contrast market at a very high cost. This paper aims at the low
caused by movement in a video image. Several cost efficient security system having user friendly
techniques for motion detection have been proposed, functional features which can also be controlled from a
among them the three widely used approaches are remote location. In addition the system can also be
background subtraction optical flow and temporal used to track the object of a predefined color rendering
differencing. Background subtraction is the most it useful for spying purposes.
commonly used approach in present systems. The
principle of this method is to use a model of the 2 HARDWARE SETUP
background and compare the current image with a
reference. In this way the foreground objects present in The proposed system comprises of two sections. The
the scene are detected. Optical flow is an transmitter section consists of a computer , RS232
approximation of the local image motion and specifies Interface, microcontroller, RF Transmitter, RF video
how much each image pixel moves between adjacent receiver. The Receiver section consists of a Mobile
images. It can achieve success of motion detection in Platform, RF receiver, microcontroller, RF camera,
the presence of camera motion or background motor driver, IR LEDs. The computer at the
changing. According to the smoothness constraint, the transmitter section which receives the visual
corresponding points in the two successive frames information from camera mounted on mobile platform
should not move more than a few pixels. For an works as control centre. Another function of control
uncertain environment, this means that the camera centre is to act as the web server that enables access to
motion or background changing should be relatively system from a remote location by using internet. The
small. Temporal differencing based on frame control centre is also responsible for transmitting the
difference, attempts to detect moving regions by necessary control signal to the mobile platform.
3 MODES OF OPERATION
Sumita Mishra is currently pursing doctoral degree in Electronics at DRML
Avadh University, India and working as a lecturer in electronics and The system can operate in four independent modes.
communication engineering department at Amity University, India
E mail: [email protected]
Prabhat Mishra is currently pursuing masters degree program in electronics 3.1 PC Controlled Mode
and communication engineering in Amity University, India
IJSER 2011
49
ISSN 2229-5518
In this mode the mobile platform is directly controlled
Fig 3. Setup for tracing mode

Fig 1. Setup for PC Controlled Mode
In this mode the system is made to follow the object
by control centre using a visual GUI program whose color information has been stored at the control
developed using Microsoft Visual Studio 6.0(Visual centre in program developed in MATLAB. Basically
Basic programming language). The user can control the program performs the image processing of the
the mobile platform after analyzing the received video. object and generates the control signals in order to
make the mobile platform to trace the object.
3.2 Internet Controlled Mode
This mode is an extension to the PC Controlled mode 3.4 Motion Detection Mode
where client-server architecture is incorporated. This
mode enables an authorized client computer to control In this mode the platform is made to focus on a
the mobile platform from a remote location via particular object whose security is our concern. The
internet. Client logs onto the control centre which mobile platform transmits the visual information of the
provides all control tools for maneuvering the mobile object to the control centre for analysis. A Program
platform. Instant images of the environment developed using MATLAB at the control centre is then
transmitted from the camera mounted on the mobile
Fig 2. Setup for Internet Controlled mode
platform are used to generate appropriate control

signals.
3.3 Tracing Mode

Fig 4. Setup for motion detection mode
used to analyze four consecutive and based on this

analysis a security alarm is raised if required.
4. DEVELOPMENT OF CONTROLLING PROGRAMS

4.1 Program for mode 1
IJSER 2011
50
ISSN 2229-5518
This program has been developed in Microsoft Visual

Studio 6.0(Visual Basic programming language).It
consists of 12 buttons, 2 checkboxes, 1 video box, 1
picture
box. These 16 buttons are configured as:
7 buttons to control the directions of the
platform.
2 checkboxes for controlling lights and night
vision respectively.
2 buttons for camera control (start, stop).
2 buttons for capturing video. Fig 6 A screen view of server program
1 button for capturing snapshot
The video box displays the video using VideoCapPro The server program (fig. 6) continuously listen to
ActiveX Control, received from the camera mounted the port and establishes connection if requested.
on the mobile platform and similarly the picture box It then transmits the video to the client using
displays the snapshot taken when the button for VideocapX control and provides access to its
capturing the snapshot is depressed. The program serial port.
transmits the control signals via serial port using When connect button is depressed, client
MSCOMM (Microsoft Common Control) component. program connects to the server whose IP address
has been specified in the IP text box at the given
4.2 Program for mode 2 port using WINSOCK (Windows Socket)
This program implements client-server component. Once the connection is established,
architecture in VB using TCP and socket
the server provides video feedback to the client.
programming. The client in turn controls the mobile platform
The client program (fig. 5) has total of 16 via internet.
buttons,1 video box, 1 picture box and 2 text
boxes which are configured as follows: 4.3 Program for mode 3
2 buttons for managing the connection It has been developed in MATLAB (ver 7.01 from
between client and server. Math Works).The captured image is analyzed pixel
1 text box to input the host IP address and by pixel. The screen is divided into four quadrants.
other for communicating port. Each pixel value is then compared with the stored
7 buttons to control the directions of the color value. On comparing these values with the
mobile platform. pixels of the captured image (fig. 7) of the object,
2 buttons for controlling lights and night those pixels are highlighted which matches with
vision respectively. the specified color. The highlighted pixels (fig. 8)
2 buttons for capturing video. indicate the direction in which the mobile platform
1 button for capturing snapshot. is to be moved. According to the quadrant of
2 buttons for camera control (start, stop). highlighted area, control signals are automatically
transmitted to the mobile platform to track the
object.
Fig 5. A screen view of client program

Fig 7 unprocessed image
IJSER 2011
51
ISSN 2229-5518
difference between four consecutive frames, however,

due to the unexpected traffic in the internet, delay in
communication between the server and the end user
may happen. This causes delays on executing
commands and programs and transmitting on-site
images.
ACKNOWLEDGMENT
The first author Sumita Mishra is grateful to Prof. B. P.

Fig. 8 Processed image showing red color Singh, Maj. Gen. K.K. Ohri, Prof. S.T.H. Abidi and
Brig. U. K. Chopra of Amity University, India for their
4.4 Program for mode 4 support during the research work.
Again, MATLAB is used for programming. The
GUI (fig. 9) has a button and one text box. The REFERENCES
button incorporates multiple functions such as
start camera, start and stop monitoring process.
[1]An Improved Motion Detection Method for Real-
Monitoring can only be closed if correct password
Time Surveillance
is entered in the text box. On activation the
Nan Lu, Jihong Wang, Q.H. Wu and Li Yang IAENG
monitoring of scene in front of camera is
International Journal of Computer Science, 35:1,
performed continuously and visual information is
IJCS_35_1_16
sent to the control centre. At the control centre [2]Robust Salient Motion Detection with Complex
image processing is performed to detect any
Background for Real-time Video Surveillance, Y.L.
motion. Tian and A. Hampapur IEEE Computer Society
Workshop on Motion and Video Computing,
4.5 Program for Mobile Platform An AVR
Breckenridge, Colorado, January 5 and 6, 2005.
ATmega8 microcontroller, L298 motor driver IC is
[3] "Real-Time Mobile Robot Teleoperation Via
embedded in the mobile platform. The controller is
Internet Based On Predictive Control", Shihua
programmed to drive the motor driver IC which in Wang,Bugong Xu, Yunnhui LIU Yeming ZHOU, Front
turn drives the motors of the mobile platform. Mech. Engg. China, 3(3):299-306 2008.
AVR studio ver 4.0 is used as a program editor,
[4] "Internet Control Of a Domestic Robot using a
WINAVR is used for compiling the code. The
Wireless Lan", Johan Potgeiter, Glen Bright, Olaf
program is written into the microcontroller using Diegel, Sylvester Tlale Mechatronics and Robotic
Ponyprog (ver 1.17) which interfaces the
Research Group Institute of Technology and
programmer with the parallel port.
Engineering, Massey University, Auckland
[5] "Controlling Robot Through Internet Using Java",
Mr. Ravindra Thamma, Dr. Luke H. Huang, Dr. Shi-Jer
Lou and Dr. C. Ray Diez, Volume 20, Number 3 - June
5. CONCLUSIONS 2004 through August 2004
[6]http://gow.epsrc.ac.uk/ViewGrant.aspx?GrantRef
The designed advanced real time video motion =EP/E027253/1
detection system allows user to maneuver the mobile [7]http://www.mathworks.com/matlabcentral/f
platform from a remote location. The system also ileexchange/4412
provides feedback to the end user in terms of visuals [8]http://www.mathworks.com/matlabcentral/newsr
rendering the system useful for spying purposes. This eader/view_thread/243698
system integrates the functionality of four modes and
selects one of them for unique application. The image
processing discussed in mode 3 proves to be useful in
industry applications and mode 4 can be useful for
monitoring the highly restricted areas. Unlike most
previous methods for real time video analysis the
suggested approach used in mode 4 utilizes the
IJSER 2011
52

ISSN 2229-5518
Handoff Analysis for UMTS Environment

Pankaj Rakheja, Dilpreet Kaur, Amanpreet Kaur
Abstract UMTS is one of the third generation mobile telecommunication technologies. It supports various multimedia applications and
services at an enhanced data rate with better security. It also supports mobile users and for that there is a process called handover
where new channels are assigned to the user when it moves from a region covered by one node to a region covered by other node. In
this paper we are analysing the effect of handover over the perform ance of the system.
Index Terms DPCH, Handover, UTRA.
place the system needs to identify the user terminal and

INTRODUCTION monitor its signal strength and setting of a threshold value
below which a call or application drops and enabling new
U niversal Mobile Telecommunications System [1-2] is a
third-generation broadband which supports packet-based
transmission of text, digitized voice and video. The
channel allocation before this level.
multimedia here can reach data rates up to 2 megabits per There is handoff margin which needs to be optimized for
second (Mbps). It also offers a consistent set of services to proper synchronization. It is the difference between signal
mobile computer and phone users, no matter where they are strength at which handover should occur and the minimum
located in the world. It is based on the Global System for required signal strength. If it is too low then there will be
Mobile Communications (GSM) standard i.e. it is overlaid on insufficient time to complete the process and if it is too large
GSM. It is also endorsed by major standard bodies and then unnecessary handovers will occur. The most important
manufacturers as the planned standard for mobile users around thing is the handovers are not visible to the users.
the world. It can ensure a better Grade of Service and Quality
of Service on roaming to both mobile and computer users. Handover types
Users will have access through a combination of terrestrial
wireless and satellite transmissions. Handovers can be broadly classified into two types namely:
Intracellular and Intercellular handover. In the Intracellular
Cellular telephone systems used previously [3] were mainly handover, mobile or user terminal moves from one cellular
circuit-switched, meaning connections were always dependent system to another. And in the Intercellular handover, user
on availability of circuits. A packet-switched connection uses terminal moves from one cell to the other. This is further
the Internet Protocol (IP) [4-5] which uses concept of virtual classified into soft and hard handover.
circuit i.e. a virtual connection is always available to connect
an endpoint to the other end point in the network. UMTS has Soft handover
made it possible to provide new services like alternative
billing methods or calling plans. For instance, users can now Here we follow make before break concept where the user
choose to pay-per-bit, pay-per-session, flat rate, or asymmetric terminal is allocated new channels first and then previous
bandwidth options. The higher bandwidth of UMTS also channels are withdrawn. The chances of losing continuity are
enabled other new services like video conferencing. It may very less but it needs user terminal or mobile to be capable of
allow the Virtual Home Environment to fully develop, where toning to two different frequencies. The complexity at user
a roaming user can have the same services to either at home, end increases a lot. It is quite reliable technique but here
in the office or in the field through a combination of channel capacity reduces.
transparent terrestrial and satellite connections.
Hard Handover
I. OVERVIEW
Here we follow break before make concept where from the
user terminal previously allocated channels are first
The term handover [6] is also known as handoff. Whenever withdrawn and then new channels are allocated. The chances
a user terminal moves into area covered by a different RNC of call termination are more than in soft handover. At the user
while the conversation is still going on, then new channels are terminal complexity is less as it need not be capable of toning
allocated to the user terminal which is now under different to two different frequencies. It provides advantage over soft
control node or MSC. This is carried out to ensure continuity handover in terms of channel capacity but it is not as reliable
of communication and to avoid call dropping. For this to take as soft handover.
IJSER 2011
53

ISSN 2229-5518
object here corresponds to one transport channel in case of

Prioritizing handoffs BLER. A measurement object corresponds to one timeslot in
case of SIR (TDD only).
Handoff requests are more important than new call requests
or application requests as call dropping in between will be UE-internal measurements: Measurements of UE
more annoying for the user than not being able to make a new transmission power and UE received signal level.
call. So a guard channel is especially reserved for the handoffs.
We also queue the requests made for proper flow and order UE positioning measurements: Measurements of UE
control. position.
The UE supports a number of measurements running in
The most obvious cause for performing a handover is that parallel. The UE also supports that each measurement is
due to its movement a user can be served in the another cell controlled and reported independently of every other
more efficiently (like less power emission, less interference measurement.
etc). It may however also be performed for other reasons
which may be system load control.
II. WORK DONE
Classification of cells
We have designed three scenarios [7] where handovers
Active Set: It is defined as the set of Node-Bs the UE is occur when the user terminal moves from the area of one node
simultaneously connected to (i.e., the UTRA cells currently to the other node. In order to enable communication between
assigning a downlink DPCH to the UE constitute the active source and destination, we made analysis for effect of speed
set). and number of handovers over throughput, average jitter,
average end to end delay etc. Here in scenario 1, shown in
Monitored Set: It is defined as the set of nodes not in the figure 1, the terrain is 1500 sqms and there are two nodes. The
active set but are included in CELL_INFO_LIST. UEs move from the area of one node to that covered by the
other node while application is still in active state, so one
Detected Set: It is defined as the set of nodes neither in the handover has occurred here; in scenario 2, shown in figure 2,
active set nor in CELL_INFO_LIST but are detected by UT the terrain is 2500 sqms and there are three nodes. The UEs
special considerations in UMTS environment. move from the area covered by one node to the second and
then to the third one while application is still in active state, so
In UMTS environment the different types of air interface two handovers have occurred here; and in the scenario 3,
measurements are: shown in figure 3, the terrain is 3500 sqms and there are four
nodes. The UEs move from the area covered by one node to
Intra-frequency measurements: Those measurements which the second to the third and then to the fourth one while
are carried out on downlink physical channels at the same application is still in active state, so three handover have
frequency as that of the active set. The measurement object occurred. Here we have taken two users one travelling at 16
here corresponds to one cell. m/s and the other at 20 m/s respectively. The later one is
called as Fast UT (User Terminal) while the former one is
Inter-frequency measurements: Those measurements which referred to as Slow UT (User Terminal).
are carried out on downlink physical channels at frequencies
that differ from the frequency of the active set. The The screenshots of three scenarios designed to analyse
measurement object here corresponds to one cell. impact of handover on the overall performance of the system
are:
Inter-RAT measurements: Those measurements which are
carried out on downlink physical channels belonging to
another radio access technology than UTRAN, e.g. GSM. The
measurement object here corresponds to one cell.
Traffic volume measurements: Those measurements which

are carried out on uplink channels to analyse the volume of
traffic on them. The measurement object here corresponds to
one cell.
Quality measurements: These measurements are carried out

on downlink channels to obtain the various quality parameters,
e.g. downlink transport block error rate. The measurement
IJSER 2011
54

ISSN 2229-5518
Throughput Fast UT Slow UT
One handover 4171 4188
Two handovers 4170 4189
Three handovers 4169 4188
Average Jitter Fast UT Slow UT
One handover 0.233 0.213
Two handovers 0.0556 0.009

Figure 3: Screenshot of scenario for three handovers
Three handovers 0.17 0.038 The values obtained by running the simulations for
throughput, average jitter and average end to end delay [7] for
various scenarios shown in figures (1-3) are:
Table 1: Throughput values for handovers
Table 2: Average jitter values for handovers
Table 3: Average end to end delay values for handovers
Average end to
Fast UT Slow UT
end delay
One handover 0.67517 0.6755
Figure 1: Screenshot of scenario for one handover

Two handovers 0.45 0.39
Three handovers 0.52 0.44
The plots drawn using these values obtained are:
Figure 2: Screenshot of scenario for two handovers
Figure 4: Graph of throughput for handovers
IJSER 2011
55

ISSN 2229-5518
[3] Oudelaar, j, Evolution towards UMTS, 5th IEEE

International Symposium on Personal, Indoor and
Mobile Radio Communications, 1994
[4] Fan, L. Sheriff, R.E. Gardiner, J.G, Satellite-UMTS
service provision using IP-based technology, 2000
IEEE 51st Vehicular Technology Conference
Proceedings.
[5] Jin Yang Kriaras, Migration to all-IP based UMTS
networks, First International Conference on 3G Mobile
Communication Technologies, 2000
[6] Hyeyeon Kwon Kyung-yul Cheon Aesoon Park,
Analysis of WLAN to UMTS Handover, 2007 IEEE
Figure 5: Graph of average jitter for handovers
66th Vehicular Technology Conference, 2007
[7] Qualnet Programmers Guide, Scalable Network
Technologies.
Figure 6: Graph of average end to end delay for handovers
III. CONCLUSION
On analysing the results obtained after running different

simulations, we can say that the system has better
performance for slow speed users as values of performance
determining parameters like throughput, average jitter and
average end to end delay are better than that of high speed
users. With number of handovers, throughput does not vary
much; the average jitter and average end to end delay first fall
down and later increases. So, overall performance of system is
good in case of handover as average jitter and average end to
end delay are not much and throughput is also good as it does
not vary much.
IV. REFERENCES
[1] Falciasecca, G. Frullone, M. Grazioso, P. Riva, G. Serra,

A.M., Performance evaluation of a UMTS in a urban
environment, 1991, Sixth International Conference on
Mobile Radio and Personal Communications
[2] van Nielen, M.J.J. ,R. PTT Nederland NV,
Leidschendam , UMTS: a third generation mobile
system, Third IEEE International Symposium on
Personal, Indoor and Mobile Radio Communications,
1992. Proceedings, PIMRC '92
IJSER 2011
56
ISSN 2229-5518
Performance and Emission Characteristics of

Stationary CI Engine with Cardnol Bio Fuel
Blends
Mallikappa, Rana Pratap Reddy, Ch.S.N.Muthy
Abstract The compression ignition engine is the most popularly used prime mover. The compression ignition (CI) engine moves a large
portion of the worlds goods & generates electricity more economically than any other device in their size range [1]. All most all the CI engines
use diesel as a fuel, but the diesel is one of the largest contributors to environmental pollution problems. The application of bio diesel as a
substitute for conventional petroleum fuel in diesel engine gain ever increasing demand throughout the world wide, because it is produced
from renewable resources, bio degradable and potential to exhaust emissions & use of bio diesel in diesel engines generates rural employ-
ment opportunities by cultivating such oil producing crops[1-5]. In this research work the detailed investigation on performance and emission
characteristics of four stroke single cylinder engine with variable loads were studied, cardnol bio fuel volumetric blends like 0, 10, 15, 20%,
and 25% were used. The results indicate that brake power increases (by 76% approximately) as load increases. Brake specific energy con-
version decreases (by 30-40 % approximately) with increase in load. Brake thermal efficiency increases with higher loads and emission levels
(HC, CO, NOX) were nominal up to 20% blends.
Key words: Compression Ignition, characteristics, cardnol bio fuel, Performance, Emissions

Nomenclature
BSEC : Brake Specific Energy Consumption
B.T.E : Brake thermal efficiency
B10 : Blend with 10%bio fuel
CBF : Cardnol Bio Fuel
CI : Compression Ignition
CO : Carbon Monoxide
DR-CNSL: Double Refined Cashew nut Shell Liquid
EGT : Exhaust Gas Temperature
HC : Hydro Carbons
IC : Internal Combustion
NOx : Nitrogen oxide
ppm : Parts per million
Cs : Centistokes
1 INTRODUCTION
I N todays world the majority of automotive and trans-

portation vehicles are powered by compression ignition
engines. The compression ignition engine moves a
Plenty of scientific articles and research activities from
around the world were printed and recorded. Oils from
coconut, soy bean, sunflower, safflower, peanut, linseed
large portion of the worlds goods & generates electricity and palm were used depending on what country they
more economically than any other device in their size grow abundantly. It has been reported that in diesel en-
range. All most all the CI engines use diesel as a fuel, but gines; vegetable oils can be used as fuel, straight as well
the diesel is one of the largest contributors to environ- as in blends with the diesel. It is evident that [2] there are
mental pollution problems. Bio fuel is an alternative to various problems associated with vegetable oils being
petroleum based fuel, renewable energy source, bio de- used as fuel in compression ignition engines, mainly
gradable and non-toxic fuel, being beneficial for reser- caused by their high viscosity. The high viscosity is due to
voirs, lakes, marine life and other environmentally sensi- the molecular mass and chemical structure of vegetable
tive places such as large cities and mines & use of bio di- oils, which in turn leads the problems in pumping, com-
esel in diesel engines generates rural employment oppor- bustion and atomization in the injector system of diesel
tunities by cultivating such oil producing crops [1-5]. engine. Due to the high viscosity, vegetable oils normally
The issue of energy security led governments and re- introduce the development of gumming, the formation of
searchers to look for alternate means of renewable and injector deposits, ring sticking as well as incompatibility
environment-friendly fuels. Bio fuel has been one of the with conventional lubricating oils in long-term opera-
promising, and economically viable alternatives. Fuel and tions.
energy crisis and the concern of society for depleting India is the largest producer, processor and exporter of
worlds non-renewable resources initiate various sectors Cashews, Anarcadium Occidentale Linn, in the world [6].
to look for alternative fuels. One of the most promising It was brought to India during the 1400 by Portuguese
fuel alternatives is the vegetable oils and their derivatives. missionary. Cashew came conquered and took deep root
IJSER 2011
57
ISSN 2229-5518
in the entire coastal region of India. While the tree is na- carboxylated oil by heating the oil to a temperature of
tive to central and Southern America it is now widely 170C to 175C under reduced pressure of 30-40 mm.
distributed throughout the tropics, particularly in many mercury. The next two steps are the same as above for the
parts of Africa and Asia. In India Cashew nut cultivation production of both cardol or cordnol and anacardol.
now covers a total area of 0.70 million hectares of land,
producing over 0.40 million metric tons of raw Cashew
nuts. The Cashew (Anacardium Occidentale) is a tree in 1.2.1Cardnol
the flowering plant family Anacardiaceae. The plant is DR-CNSL - Double Refined Cashew nut Shell Liquid. The
native to northeastern Brazil, where it is called by its Por- Cashew Nut Shell Liquid (CNSL) obtained by pyrolysis.
tuguese name Caju (the fruit) or Cajueiro (the tree). It is It mainly consists two naturally produced phenolic com-
now widely grown in tropical climates for its cashew pounds: Anacardic acid 90% Cardol or cardnol 10%.
"nuts" and cashew apples.
1.1 Specification of Cashew nut shell
The shell is about 0.3 cm thick, having a soft feathery out-
er skin and a thin hard inner skin. Between these skins is
the honeycomb structure containing the phenolic material
known as CNSL. Inside the shell is the kernel wrapped in
a thin skin known as the teesta.
1.2 Composition of cashew nut Cardnol obtained by pyrolysis from dr-csnl oil was uti-
The shell is about 0.3 cm thick, having a soft feathery out- lized for testing purposes. Cardnol is a naturally occur-
er skin and a thin hard inner skin. Between these skins is ring phenol manufactured from cnsl. It is a monohydrox-
the honeycomb structure containing the phenolic material yl phenol having a long hydrocarbon chain in the Meta
known as CNSL. Inside the shell is the kernel wrapped in position.
a thin skin known as the testa.The nut consists of the fol- C6H4 (OH)-(CH2)7-CH=CH-CH2-CH=CH (CH2)2 -CH3
lowing kernel 20 to 25%, kernel liquid 20 to 25%, testa 2%,
others rest being the shell. The raw material for the manu-
facture of CNSL is the Cashew.
According to the invention [6] CNSL is subjected to
fractional distillation at 200 to 240C under reduced
pressure not exceeding 5mm. mercury in the shortest
possible time which gives a distillate containing cardol Reason for using the cardnol as alternative fuel: - it
and the residual tarry matter, for example, in the case of a is renewable, it is cost effective, easily produced inexpen-
small quantity of oil, say 200 ml/ the distillation period is sively in most regions of the world, results in reduced [up
about 10 to 15 minutes. A semi-commercial or commercial to certain extent] emissions compared with petro-diesels,
scale distillation of CNSL may however take longer times. results in no detrimental effects to the engine, non edible
It has been found that there are certain difficulties of op- and it is extracted from the cashew nut shell not from the
eration with regard to single-stage fractional distillation seed.
method, i.e. frothing of the oil which renders difficult the 2 EXPERIMENTAL
fractionation of cardol and also formation of polymerised The main objective was to study the performance and
resin. These difficulties can be over come in the two-stage emission characteristics of the CI engine when Cardnol
distillation, if care is taken not to prolong the heating; this and pure diesel volumetric blends were used and also to
is to avoid the undue formation of polymerised resins investigate which combination of fuel blend is suitable
and possible destruction partially or completely of the for diesel engine at all load conditions from both perfor-
cardol or anacardol. When CNSL is distilled at a reduced mance and emission point of view. Experimentation has
pressure of about 2 to 2.5 mm. mercury, the distillate con- been conducted up to cardnol bio fuel volumetric blends
taining anacardol and cardol distils firstly at about 200C like 0, 10, 15, 20%, and 25%, because the viscosities (refer
to 240C. This first distillate is then subjected to a second table 1 for properties of cardnol bio fuel blends) of higher
distillation under the same identical conditions of tem- blends are more than the international standard limits
perature and pressure when the anacardol distils over at [ASTM-allowable limits only up to 4-5 centistokes].
a temperature of 205C to 210C and the cardol distils
over at a temperature of 230C to 235C. In practice it has
been found that the preliminary decarboxylation of the oil Properties Diesel B10 B15 B20 B25 B30
is essential, since there will be excessive frothing, which
renders the distillation procedure unproductive and un- Flash point (C) 50 53 55 56 58 61
economical. A specific feature of this invention is that
Density(Kg/m3) 817 823 829 836 841 846
both cardol and anacardol may be obtained by a three-
step process. The first step of the process is to get the de-
Viscosity at 2 2.5 3.1 3.5 4.2 5.5
IJSER 2011400C
(Centistokes)
Calorific value 40000 40130 40196 40261 40326 40392
(KJ/Kg)
58
ISSN 2229-5518
2.2 Properties surements.

Following table 1 indicates the properties of cardnol bio
fuel blends. Lower Calorific value of the diesel has been
considered for calculations. Table 1 Properties
2.3 Transesterification
Selection of raw materials: cardnol oil sample, anhydrous
methyl alcohol 99% grade laboratory reagent type, So-
dium Hydroxide was selected as the catalyst.
2.3.1 Procedure
About 4 grams of Naoh (catalyst) is dissolved in 200 ml
methanol to prepare alkoxide, which is required to acti-
vate the alcohol. Then stirring is done vigorously in a
covered container until the alkali is dissolved completely
for twenty minutes. Mixture is protected from atmos-
phere carbon dioxide and moisture as both destroy the Fig 1 Experimental setup
catalyst. The alcohol catalyst (Naoh) mixture is then In this investigation the various performance and emis-
transferred to the reactor containing 700 ml moisture free sion tests were conducted on four strokes single cylinder
crude cardnol oil. Stirring of the mixture is continued for engine manufactured by M/s Kirloskar (as shown in
90 minutes at a temperature between 60-65 degrees. The fig 1) Company limited. The parameter involved in per-
round bottom flask was connected to a reactor condenser formance analysis has been measured using the engine
and the mixture was heated for approximately three software supplied by the manufacturer.
hours. 2.4.1 Specifications of the engine
Name of the engine: KIRLOSKAR, TV1
General details: 4 stroke, C.I, Vertical
2.3.2 Inference and observation Type of cooling: Water cooled
The mixture was distilling and condensing within the Number of cylinders: 1
reactor Condenser, no glycerin, because CNSL is ex- Bore: 87.5 mm
tracted from honeycomb structure (shell) of a cashew nut. Stroke: 110mm
The color of cardnol oil slightly changed from dark Rated power: 5.2 B.H.P at 1500 rpm
brown to light brown color and an average of 95% recov- Dynamometer: Eddy current dynamometer
ery of bio fuel was possible. Compression ratio: 12:1 to 17.5:1
2.4 Experimental procedure 3. RESULTS AND DISCUSSIONS

The experiments were conducted on a direct injection
The tests were conducted up to 25% blends, because the compression ignition engine for various loads with an
viscosity of above 25% blends exceeds the international intention of studying the behavior of the engine in regard
standard limits (i.e. more than 5 Cs). The load test was to various emissions, and performance characteristics
conducted for different loads i.e. no load, when it was run on different volumetric blends and the
25%load,50%load, 75%load and full load conditions and results of the performance test and the emission studies
for blends such as 0%, 10%, 15%, 20%&25%of Cardnol. conducted on the engine are plotted in the following
The Orotech exhaust gas analyzer used for emission mea- (Characteristics graphs) figures.
3.1 Brake specific Energy consumption
IJSER 2011
59
ISSN 2229-5518
esel. This lower brake thermal efficiency obtained

could be due to lower calorific value and increase in
fuel consumption as compared to B20.
3.3 .Exhaust gas temperature & NOx Emissions
Exhaust gas temp.v/s Load
800
N O x E m iss io n s i n P P m
700
B10Texh,NOx
600
500 B15DTexh,NOx
400 B20Texh,NOx
300 B25Texh,NOx
200
Dtexh,NOx
100
0
Fig 2.Brake specific Energy consumption 0 100 200 300 400 500 600
Temperture in oC
(*103Kj/Kw-hr) v/s load
Fig2.Depicts that, the brake specific energy consumption
decreases by 30 to 40% approximately with increases in
load conditions. This reverse trend was observed due to Fig 4.Exhaust gas temp. v/s NOx Emissions
lower calorific value with increase in bio fuel percentage
in the blends. The variations of exhaust gas temperature and Nox emis-
sions with respect to engine loading are presented in the
in fig.4.The exhaust gas temperature increases linearly
form 180o C at no load to 480 o C at full load conditions.
This increasing trend of EGT is mainly because of gene-
rating more power and consumptions of more fuel at
higher loads.
3.2 Brake thermal efficiency

3.4 HC Emissions
Load v/s HC Emissions
40
35
HC Emissions in %
30
B10
25
B15
20
B20
15
B25
10
5
0
0 5 10 15 20
Load in N-m
Fig 3.Brake thermal efficiency v/s Load

Fig 5. HC Emissions
The variation of brake thermal efficiency with load for
different volumetric blends is presented in fig.3.In all cas- From the figure 5 it has been observed that HC emissions
es, it increased with increase in load. This was due to re- are nominal up to B20, and more at B25, the reason for
duction in heat losses and increase in brake power with this may be incomplete combustion.
increase in load. The maximum thermal efficiency for B20
(31%) was higher than that of the diesel. The brake ther- 3.5 Carbon Monoxide Emissions
mal efficiency obtained for B25 was less than that of di-
IJSER 2011
60
ISSN 2229-5518
The Nox emissions (ppm) increases with increasedpro-

portion of blends and also with higher EGT. This trend
mainly because of presence oxygen in bio fuel, this leads
to more oxidation at higher temperature and responsible
for more Nox emissions.
The HC emissions are nominal up to B20,and more at
B25,the reason for this is the incomplete combustion.
The Carbon monoxide emissions increases with higer
blends, and increases slightly more after 20% blends. The
minimum and maximum Co produced was 0.03-0.08%.At
higherloadsCo emissions slightly decreased. At elevated
temperature, performance of the engine improved with
relatively better burning of the fuel resulting in decreased
Co.
Fig 6. Carbon Monoxide Emissions
From this investigation it has been observed that up
Carbon monoxide emissions increases with higher blends, to20% blends of cardnol bio fuels may
and increases slightly more after 20% blends. The mini- be used in CI engines without any modifications.
mum and maximum Co produced was 0.03-0.08%.
At higher loads Co emissions slightly decreased. At ele-
vated temperature, performance of the engine improved ACKNOWLEDGEMENTS
with relatively better burning of the fuel resulting in de-
creased Co. We are very much thankful to M/s Bangalore test house,
Bangalore for testing various properties of cardnlo oil, &
Dr.Sharanappa, Asst.Proff.of Mech.Engg.Dept, and
Mr.Viswanath, Instructor R.I.T.M Bangalore, for their
kind help during experimentation work. And special
thanks Mr.Ganesh kamath of M/s Sanur Cashew nut In-
dustries, Karkala for supplying CNSL to our experimen-
tation work.
CONCLUSIONS
The cnsl and its extracts showed promising results in
terms of engine performance in par with conventional CI
engine fuels. Based on the results of the study the follow- REFERENCES
ing conclusions were drawn.
[1] Alan C, Lloyd B, Thomas A. Cackette. Diesel Engines:
The significant factor of cardnol bio fuel is its low
Environmental Impact and Control California Air
cost, its abundance and it is a byproduct of
Resources Board, Sacramento, California ISSN1047-
cashew nut industries.
3289 J. Air &Waste Manage. Assoc. 51:809-847
The brake specific energy consumption decreases by 30 [2] Ayhan Demirbas Studies on biodiesel from vegetable
to 40% approximately with increases in load conditions. oils via transesterifications in supercritical
This reverse trend was observed dueto lower calorific Methanol Energy Conversion and Management 44
value with increase in bio fuel percentage in the blends. (2003) 20932109
[3] Fernando Netoda Silvaa,*, Ant_onio Salgado Pratab,
The brake thermal efficiency increases with higher
Jorge Rocha Teixeiraca Technical feasibilityassess
loads. In all cases, it increased with increase in load. This
ment of oleic sunflower methyl ester utilization in
was due to reduction in heat losses and increase in brake
diesel bus engines. Energy conversion and manage
power with increase in load.The maximum thermal effi-
ment 44 (2003) 2857-2878
ciency for B20 (31%) was higher than that of the diesel.
[4] K.Pramanik Properties and use of jatropha curcas oil
The brakethermal efficiency obtained for B25 was less and diesel fuel blends In CI engine, Renewable Ener
than that of diesel. This lower brake thermal efficiency gy 28 (2003) 239248
obtained could be due to lower calorific value and in- [5] N. Stalin and H .J. Prabhu Performance test of IC en
crease in fuel consumption as compared to B20. gine using Karanja bio diesel blendingwith diesel
ARPN Journal of engineering and applied science
vol.2, no5, October 2007
IJSER 2011
61
ISSN 2229-5518
[6] Piyali Das, T.Sreelatha, Anurada Ganesh Bio oil from

pyrolysis of cashew nut shell-characterization and
related properties Biomass and bio energy
IJSER 2011
62
ISSN 2229-5518
Morphological Space And Transform Systems

Ramkumar P.B, Pramod K.V
Abstract Mathematical Morphology in its original form is a set theoretical approach to image analysis.It studies image
transformations with a simple geometrical interpretation and their algebraic decomposition and synthesis in terms of elementary
set operations.Mathematical Morphology has taken concepts and tools from different branches of Mathematics like algebra
(lattice theory) ,Topology,Discrete geometry ,Integral Geometry,Geometrical Probability,Partial Differential Equations etc.In this
paper ,a generalization of Morphological terms is introduced.In connection with algebraic generalization,Morphological
operators can easily be defined by using this structure.This can provide information about operators and other tools within the
system
Index Termsmorphological space, transform systems, slope transforms, legendre, kernel.
1 GENERALIZED STRUCTURE FOR MATHEMATICAL MORPHOLOGY
1.1 Definition: Morphogenetic field

W ,
L ET X and W P (X ) such that i) , X Exampl 3.If
,., )
(V, ) (V, ) are groups then (V,
is called a bounded lattice ordered group or

ii) If B W then its complement B W iii) If
Bi W is a sequence of signals defined in X, then blog.[11]
Proposition 1. Clodum is an operator space.

Bi W. Since Clodum is a particular case as mentioned above ,
n 1
we can consider it as an operator space.
Let A= : W U / ( Ai ) ( Ai ) & ( Ai ) ( Ai ) .Then WU is
Proposition 2. Blog is an operator space.
called Morphogenetic field [14] where the family Wu is Blog is another example for operator space. Similar par-
the set of all image signals defined on the continuous or ticular cases exist corresponding to the algebra or geome-
discrete images Plane X and taking values in a set U .The try under consideration.
pair ( Wu, A ) is called an operator space where A is the 1.3 Definition:Dilation

Let be a complete lattice, with infimum and mini-
collection of operators defined on X.
mum symbolized by and respectively.[1]
1.2 Definition : Morphological Space A dilation is any operator
The triplet ( X, Wu, A ) consisting of a set X, a morphoge- that distributes over the supremum and pre-
serves the least element. ,
netic field Wu and an operator A(or collection of opera-
1.4 Definition: Erosion
tors) defined on X is called a Morphological space.[14] An erosion is any operator that distributes
Example 1.If X = Z then it is called Discrete Morphologi-
2 over the infimum [1]. = , =U
cal space 1.5 Definition:Morphological Adjunctions

Example 2.Let V be a complete lattice. Let ( X ,Wu , A) & (Y ,Wu , A) be a morphological spaces.
If X= V and A= (V, ,., ) where , are dilation &
The pair ( A, A) is called an adjunction iff
erosion then (V, ,., ) becomes a commutative

complete lattice ordered double monoid or Clodum [11] A( X ) Y X A(Y ) where A is an inverse opera-
where (V, ) (V, ) are commutative monoids. tor of A.
IJSER 2011
63
ISSN 2229-5518
Proposition 3.Let ( X ,Wu , ) & (Y ,Wu , ) be a morpholog- processed by erosion systems 2) A single valued slope
transform for signals processed by dilation systems 3) A
ical spaces with operators dilation and erosion on A. Then
multi valued transform that results by replacing the su-
( X ) Y X (Y ) .
prema and infima of signals with the signal values at sta-

Proposition 4(for lattice). Let ( X ,Wu , A) & (Y ,Wu , A) be tionary points.

a morphological spaces. The pair ( A, A) is called an adjunc- 2.2 Special Case Continuous Time Signals
tion iff u , v X , an adjunction (lu , v , mv , u ) on U such All the three transforms stated above coincide for conti-
nuous-time signals that are convex or concave and have

that A( x (u )) mv , u ( x(v)) and an invertible derivative. This become equal to the Legen-
v X
A( y (v)) l u , v ( y (u )) , u , v X , x, y WU .[1] dre transform (irrespective of the difference due to the

u X
boundary conditions).
1.6 Definition:Morphological Kernel
The operator defines a closure called morpho- 2.3 Morphological Signal Operators
The morphological signal operators are defined by using
logical closure and defines a kernel, called
Lattice Dilation and Erosion of Signals. The morphologi-
morphological kernel.
cal signal operators are parallel or serial inter connections
of morphological dilation and erosions, respectively, de-

2 SLOPE TRANSFORMS - GENERALIZATION
fined as
2.1 Introduction
Fourier transforms are most useful linear signal transfor- ( f g )( x) d f ( x y ) g ( y )
yR
(1)
mations for quantifying the frequency content of signals and ( f g )( x) d f ( x y ) g ( y )
yR
and for analyzing their processing by linear time inva-
Where denotes supremum and denotes infimum.
riant systems .They enable the analysis and design of li-
near time invariant systems (LTI)in the frequency do- 2.4 Legendre Transforms
main. Let the signal x(t ) be concave and assume that there
Slope transforms are a special type of non linear signal dx

exist an invertible derivative x . Imagine that the
transforms that can quantify the slope content of signals. dt
It provide a transform domain for morphological systems. graph of x,not as a set of points (t , x(t )) but as the low-
They are based on eigen functions of morphological sys- er envelope of all its tangent lines . The Legendre trans-
tems that are lines parameterized by their slope. Dilation form [12] of x is based on this concept .The tangent at a
and Erosions are the fundamental operators in Mathemat- point (t , x(t )) on the graph has slope and intercept
ical Morphology. These operators are defined on lattice equal to X x (t ) (t )
algebraic structure also. Based on this, Slope transforms
X L ( ) x[( x ) 1 ( )] ( x ) 1 ( ) where f 1
de-
are generally divided into three.
notes the inverse.
They are 1) A single valued slope transform for signals
IJSER 2011
64
ISSN 2229-5518
The function XL of the tangents intercept versus the 3 RESULTS BASED ON THE GENERALIZED
STRUCTURE
slope is the Legendre transform of x [12]
3.1 Definition:Self Conjugate Operator Space

and x(t ) X L [( X L ) 1 (t )) t ( X L ) 1 (t )] If the An operator space (Wu, A) is called self conjugate if it has
signal x is convex, then the signal is viewed as the upper a negation.
envelope of its tangent lines. Example4. A clodum V has conjugate a * for every a

2.5 Definition:Upper Slope Transform such that (avb)* = a* b* and (a b) = (a* * b*) [11]
Example5. If V is a blog [4] then it becomes self conjugate
For any signal x:R R its upper slope transform [12]
by setting

is the function X : R R with a 1 , whenV inf a V sup
a*=
inf a
[11]
V sup, whenV
X ( ) x (t ) t , R .The mapping between V inf, whenV
sup a
tR
the signal and its transform is denoted by A : x X .

Example6. If X is a concave class then A* x (t)= x(-t) where
If there is one to one correspondence between the signal
A* = A (A ) .
A
and its transform, then it is denoted by x(t ) X ( ) . 3.2 Definition:Self Conjugate Morphological Space
Proposition6. Upper Slope Transform of a concave signal is If the operator space (Wu, A) is self conjugate then the
equal to its Legendre Transform. morphological space (X, Wu, A) is called a self conjugate
Proof.Let x(t ) be a concave signal .Let it has an invertible morphological space.
derivative. For each real , the intercept of the line pass- 3.3 Definition:Operatable Functions
ing from the point (t , x(t )) in the signals graph with Let (X, Wu, A) be a morphological space. The collection K
(X, W, A) of operatable functions consists of all real va-

slope is given by x(t ) t .
lued morphologically operatable functions x(t) defined on
For a fixed , assume that t varies. Let there be a time
X such that x(t) has finite operatability with respect to A.

instant t for which the intercept attains its maximum
A morphologically operatable function x K iff
value. The intercept attains its maximum value when the
x K .ie. iff A( x( )) A x( )
line becomes tangent to the graph.Therefore x (t )
. 3.4 Definition:Morphological Transform Systems
Let ( X, Wu, A) be a perfect morphological space and K=
Corresponding to the change in , the tangent also
K ( X,Wu, A) be an operatable space. K is called a mor-
changes,and the maximum intercept becomes a function
phological transform system if
of the slope .By its definition ,the upper slope trans-
form [12] is equal to this maximum intercept function. AxT (t ) X ( ) T ( )
Thus, if the signal x(t ) is concave and has an invertible Remark1.Since K is an operatable space,
derivative,then the upper slope transform is equal to its 1) Ax(t ) y(t ) X ( ) Y ( )

Legendre transform.Hence the proof. 2) AxT (t ) X ( ) T ( )
IJSER 2011
65
ISSN 2229-5518
3.5 Definition:Morphological Slope Transform phological dilation) into an addition .This is similar to the
System
concept in Fourier transform transforms. In Fourier trans-
If A= Av in the previous definition,then K is called a
Morphological slope transform system where Av is the form a linear convolution changed into a multiplication.
Difference between the Fourier transform and its mor-
upper slope transform.
Let (X, Wu, A) be a self conjugate morphological space. If phological counterpart, the slope transform is that the
X is a concave class then A * (x(t) )= x(-t) where A*= A (A Fourier transform is invertible but the slope transform
only has an adjoint. In the sense of adjunctions, this

v) and A is the lower slope transform. Also Av ( xc )
means that the inverse of the slope transformed signal
= A
c
( xc )
is not the original signal but only an approximation with-
Proposition 7(Characterization of Slope Transforms). in the sub collection. In this paper we made an attempt
A Slope transform is an extended real valued function Av (or for generalizing the algebraic structures related to the
A ) defined on a Morphogenetic field Wu such that theory of Signal processing using Mathematical Mor-
1. Av ( )= 0 phology. Morphological operators can be redefined by
using these structures. We hope that this will be helpful

2. Av (xc) 0 xc Wu
for finding new applications in Mathematical Morpholo-
3. Av is countably additive in the sense that if (xc)is any
gy.
disjoint sequence[or sampling Signal]then Av ( xc )
REFERENCES
= A
c
( xc )
[1] John Goustias and Henk J.A.M Heijmans ,Mathematical Mor-
phology, , I.O.S Press.
Remark 2.Av takes i.e Av ( xc ) = if x(t) = [2] H.J.A.M Heijmans, Morphological Image Operators,Boston,
M.A Academic,1994 .
[3] J .Serra, Image Analysis and Mathematical Morphology,New
Av ( ) > , unless x(t) = , t t
York Academic ,1982.
[4] P .Maragos and R.W Schafer, Morphological system for multi
If x= - then Av = dimensional signal processing Proc. IEEE,Vol,78,P.D 690-
Proposition 8. Let K be a morphological transform system. 710,April 1990.
[5] P .Maragos,A representation theory for morphological image
Let X be a class of concave functions. Let x( ) X with each and signal processing. IEEE Transactions on Pattern analysis
and machine intelligence 11,(1989),586-599.
x( ) has an invertible derivative.
[6] The Matheron Representation Theorem for Gray Scale Morpho-
logical Operators, G. CROMBEZ,Proceedings of the American
Then A ( x( )) L( x( )) where L is the Legendre trans-
Mathematical Society Volume 108, Number 3, March
1990(Proceedings)
form and Av is the upper slope transform.
[7] Heijmans, H.J.A.M ,and Ronse,C.The algebraic basis of Mathe-
matical Morphology Part I, Dilations and Erosions ,Computer
vision, Graphics and Image Processing,50(1990) 245-295.
4 CONCLUSION [8] Rein Van Den Boomgaard and Henk Heijmans, Morphological
The slope transform has emerged as a transform which scale space- operators.
[9] Javier Vidal& Jose Crespo,Sets Matching in Binary Images
has similar properties with respect to morphological sig- Using Mathematical Morphology,International Conference of
the Chilean Computer Science Society.
nal processing. Fourier transform does this with respect [10] Jean Cousty, Laurent Najman and Jean Serra,Some morpholog-
ical operators in graph spaces,ISSM-2009.
to linear signal processing .Main property of slope trans- [11] Petros Maragos,Lattice Image Processing: A Unification of
Morphological and Fuzzy Algebraic Systems,Journal of
form is that it transforms a supremal convolution ( mor-
Mathematical Imaging and Vision 22:333-353,2005.
IJSER 2011
66
ISSN 2229-5518
[12] Chu-Song Chen, Yi-Ping Hung,Theoretical Aspects of Vertically

Invariant Gray-Level Morphological Operators and Their Ap-
plication on Adaptive Signal and Image Filtering, IEEE transac-
tions on signal proce ssing, vol. 47, no. 4, april 1999 1049.
[13] Petros Maragos,Slope Transforms:Theory and Application to
Nonlinear Signal Processing,IEEE Transactions on Signal Proc-
essing,Vol 43,No.4,April 1995.
[14] K.V Pramod, Ramkumar P.B , Convex Geometry and Mathe-
matical Morphology, International Journal of Computer Appli-
cations,Vol:8,Page 40-45.

Ramkumar P.B is working as Assistant Professor in Mathematics at Raja-
giri School of Engineering& Technology , Mahatma Gandhi University,
India, PH-04842432058. E-mail: [email protected]
Pramod K.V is working as Professor at Department of Computer Applica-
tions,Cochin University, India, PH-01123456789.
E-mail :[email protected]
IJSER 2011
67
ISSN 2229-5518
Bus Proximity Indicator

(An Intelligent Bus Stop)
Prof. A.P. Thakare, Mr. Vinod H. Yadav
Abstract It is always a good idea that a bus commuter waiting at a stop gets to know how far a bus is. If his route of travel happens
to be common for more than one bus- route number, it is even better for him to know which is the nearest bus or the earliest arriving
bus. This will enable him to opt for the bus or some other mode of commuting. This becomes very useful for the physically challenged
commuter, as after knowing in advance the bus arrival s/he will be ready to accommodate in the bus.
A thought of project Bus Proximity Indicator is the best solution for the above situation and is best suitable for the B.E.S.T.
(The Brihanmumbai Electric Supply & Transport) In this a wireless RF linkage between a certain bus and a bus stop can be
used for determination of the bus proximity that helps commuter to know how far his bus is. This project tells him the Bus
number, bus name and the approaching time by displaying it on the LCD which is on the bus stop. This project also satisfies
the need of automization in bus services.
Index Terms Amplitude Shift Keying, Atmels AT89C52 Microcontroller, RF encoder/ decoder IC ST12CODEC, C51 Cross
Compiler, Radio frequency transmitter, Timer astable multivibrator.
1. INTRODUCTION An
T HE Bus Proximity Indicator presented in the section uses

radio frequency of 433 MHZ. The prefixed code of the bus
is generated by the encoder ST 12 CODEC. This code is trans-
TAMV ENCODER RF TX
555
mitted after Amplitude shift Keying. The receiver positioned
at the bus stop detects the radio frequency signal and the bus
identification is done by decoder ST 12 CODEC.
2 DESCRIPTION BATTERY
The block diagram and relevant description of the same is giv-

en including of Transmitter and receiving section
2.1 Transmitter Section

The basic block diagram for the Transmitter section is as shown in the block diagram. It consists of the following
blocks:
a) TAMV 555
b) Encoder
c) RF Transmitter
d)Battery
LEGEND:
TAMV Timer astable multivibrator,
RF TX Radio frequency transmitter
a) TAMV 555:
The 555 timer IC is used as an astable multivibrator and as an address setter for triggering an IC ST12CODEC which is
used as an encoder
(Figure 1: Transmitter Section of Bus Proximity Indicator)
b) RF Encoder:
A logic circuit that produces coded binary outputs from encoded inputs. This uses ST CODEC 12BT for encoding the
data. The encoder encodes the data and sends it to RF Transmitter. The IC ST12 CODEC is a single chip telemetry de-
IJSER 2011
68
ISSN 2229-5518
vice, which may be an encoder or a decoder. When combined with a Radio transmitter / receiver it may be used to pro-
vide encryption standard for data communication system The IC ST12CODEC performs all the necessary data manipu-
lation and encryption for an optimum range reliable radio link.
Transmitter and receiver use same IC ST12 CODEC in RF encoder mode for serial communication. This IC is
capable of transmitting 12 bits containing 4 bit address bit and 8 bit data. The transmitted information is sent by RF
with 434 MHZ RF transmitter. ST12 CODEC works on 5v.
RF Transmitter:
RF transmitters uses ASK (Amplitude Shift Keying) for modulating the data send by ST12 CODEC .This modulated
information is then transmitted with 433 MHz frequency through RF antenna to receiver. It helps in transmitting data
present in encoder via antenna at particular frequency.
c) Battery:
A single 9V battery is used to supply power to the transmitter section.
2.2 Receiver Section

The basic block diagram for the Receiver section is as shown above. It consists of the following blocks,
a) RF Receiver
b) RF Decoder
c) Microcontroller
d)Power supply
e) LCD
LEGEND:
RF RX: Radio frequency receiver
LCD: Liquid crystal display
RFDC: RF Decoder
C: Microcontroller AT 89C51
(Figure 2: Receiver Section of Bus Proximity Indicator)

a) RF Receiver:
It is enhanced single chip IC RWS 434 which receives the 433.92 MHz transmitted signal, transmitted by RF transmit-
ter. It uses ASK (Amplitude Shift Keying) conventional heterodyne receiver IC for remote wireless applications.
b) RF Decoder:
A logic circuit that used to decode coded binary word. This uses IC ST12 CODEC for decoding the data which is
transmitted by IC RWS 434. The decoder converts the serial data which has been sent from RF receiver to parallel form
and sends it to microcontroller. The coded data decoded by this block is given to LCD.
c) Microcontroller (IC 89C52):

This is the most important block of the entire system. The microcontroller works at crystal frequency of 11.0592 MHz.
It receives the parallel data from ST12 CODEC IC and compares it with the program code which already stored in it.
This microcontroller has the baud rate 9600 bits/sec.
The 89C52 is a low power, high performance CMOS 8 bit microcomputer with 8k bytes of flash programmable and
erasable read only memory (PEROM).The device is manufactured using Atmels high density nonvolatile memory
technology and it is compatible with the industry standard 89C51 and 89C52 instruction set and pin out.
The on chip Flash allows the program memory to be reprogrammed in system or by a conventional nonvolatile memo-
ry programmer. By combining a versatile 8-bit CPU with flash on a monolithic chip, the Atmels AT89C52 is a powerful
microcomputer which provides a highly flexible and cost effective solution to many embedded control applications.
IJSER 2011
69
ISSN 2229-5518
d) Power Supply:
The performance of the master box depends on the proper functioning of the power supply unit. The power supply
converts not only A.C into D.C, but also provides output voltage of 5V, 1 amp. The essential components of the power
supply are Transformer, four diodes which forms bridge rectifier, capacitor which work as a filter and positive voltage
regulator IC 7805. It provides 5v to each block of the transmitter.
e) 16 X 2 LCD:
LCD modules are useful for displaying the information from a system.
These modules are of two types, Text LCD and Graphical LCD. In this project a Text LCD of size (16 x 2) with a two
line by sixteen character display is used to display the various sequence of operations during the operation of the
project. This is used for visual information purpose. The LCD will display the data coming from normal keyboard or
form microcontroller as a visual indication.
3 SOFTWARE TOOLS
A
(Figure 3:
c LCD Software tool of

RFDC
RF RX 16X2 Bus Proximity
Indicator)
KEIL
POWER SUPPLY Introduction to
C51 Cross
Compiler
The Keil C51 Cross Compiler is an ANSI C Compiler that is
written specifically to generate fast, compact code for the
8051 microcontroller family. The C51 Compi- ler generates
object code that matches the efficiency and speed of assem-
bly programming.
Software Development Cycle in KEIL
When we use Keil software tools, the project devel-
opment cycle is roughly the same as it is for any other soft-
ware development project.
1. Create a project, select the target chip from the device
database, and configure the toll settings.
2. Create source file in C or assembly.
3. Build your application with the project manager.
4. Correct errors in the source files.
5. Test the linked application.
A block diagram of the complete 8051 tool set may best illustrate the development cycle.
As demonstrated in this document, the numbering for sections upper case Arabic numerals, then upper case Arabic
numerals, separated by periods. Initial paragraphs after the section title are not indented. Only the initial, introductory
paragraph has a drop cap.
4 RESULTS AND CONCLUSION

The paper Bus Proximity Indicator exhibits the arrival of a particular bus on the display provided at the bus stop.
The intention of presenting the paper is to facilitate the commuters waiting at the city bus stops. The CODEC used in
the paper generates eight bit coded information allowing the identification up to 256 bus routes and is compatible to
the AT89C51. In addition, it works on a power supply ranging from 2 to 5 volts which makes it handy in the mobile
IJSER 2011
70
ISSN 2229-5518
bus. The transmitter and receiver used works on 434MHz at 2 12 volt and hence have dual advantage of power sav-
ing as well as a range of around 500 feet. The 500 feet (150 Meter) is quite a high range for the detection of the city bus
arrival
REFERENCES
[1] 8051 Microcontroller Architecture Programming and Application by Kenneth J Ayala
[2] 8051 Microcontroller and Embedded systems by Mazidi and Mazidi
[3] Embedded Controller Forth for the 8051 Family
[4] 8051 Microcontroller, The: Hardware, Software and Interfacing & Applications
[5] WWW. Google.Com
[6] WWW.datasheetcatlog.com
[7] digikey.com/1/parts/638617
[8] http://www.sunrom.com
[9] http://www.sparkfun.com/products/8946

Prof. A.P. Thakare Head of Department of Electronics & Telecommunication Sipnas College of Engineering & Technology Amravati 444701 Maharash-
tra India Email:- [email protected]
Mr. Vinod H. Yadav is currently pursuing masters degree program in Digital Electronics engineering in Sant Gadgebaba Amravati University,
Amravati 444701 Maharashtra India, E-mail: [email protected], [email protected], [email protected]
IJSER 2011
71
ISSN 2229-5518
Designing Aspects of Artificial Neural Network

Controller
Navita Sajwan, Kumar Rajesh
Abstract In this paper important fundamental steps in applying artificial neural network in the design of intelligent control
systems is discussed. Architecture including single layered and multi layered of neural networks are examined for controls
applications. The importance of different learning algorithms for both linear and nonlinear neural networks is developed. The
problem of generalization of the neural networks in control systems together with some possible solutions are also included.
Index Terms Artificial, neural network, adaline algorithm, levenberg gradient, forward propagation, backward propagation,
weight update algorithm.
1 INTRODUCTION
T he field of intelligent controls has become important

due to the development in computing speed, power
and affordability. Neural network based control
1.
2.
Randomly choose the value of weights in the
range -1 to 1.
While stopping condition is false,follow steps 3.
system design has become an important aspect of 3. For each bipolar training pair S:t, do step 4-7.
intelligent control as it can replace mathematical models. 4. Select activations to the input units. X0=1,
It is a distinctive computational paradigm to learn linear xi=si(i=1,2..n).
or nonlinear mapping from a priori data and knowledge. 5. Calculate net input or y.
The models are developed using computer, the control 6. update the bias and weights.
design produces controllers,that can be implemented W0=w0(old)+alpha(t-y)
online.The paper includes both the nonlinear multi-layer Wnew=wi(old)+alpha(t-y)xi.
feed-forward architecture and the linear single-layer
architecture of artificial neural networks for application in 7. If the largest weight change that occurred in step
control system design. In the nonlinear multi-layer feed- 3 is smaller than a specified value, stop else
forward case, the two major problems are the long continue.
training process and the poor generalization. To
overcome these problems, a number of data analysis X1 X w1
strategies before training and several improvement
generalization techniques are used. 0 or 1
X
2 ARCHITECTURE IN NEURAL NETWOKS

X
y
Depending upon the nature of the problems, design of
neural network architecture is selected. There are many
commonly used neural network architectures for control Wn
X
system applications such as Perceptron Network, Adaline
network, feed forward neural network. Error=t-y t
(a)ADALINE Architecture: X
ADALINE( For ADAptive LINear combiner) is a device Xn Adaptive Output error
and a new, powerful learning rule called the widrow- weights generator
Hoff learning rule this is shown in figure1.The rule
minimized the summed square error during training
involving pattern classification. Early applications of
ADALINE and its extension to MADALINE( for many Figure 1:ADALINE Neuron Model
ADALINES) include pattern recognition, weather
forecasting and adaptive control (b)Feed-forward Neural Network Architecture: It is a
important architecture due to its non-parametric, non-
linear mapping between input and output. Multilayer
feed-forward neural networks employing sigoidal hidden
ADALINE algorithm: unit activations are known as universal approximators.
IJSER 2011
72
ISSN 2229-5518
These function can approximate unknown function and I1 a1 V11 b1 w11 c1

its derivative. The feed-forward neural networks include a1
one or more layers of hidden units between the input and V13 V12 w12 Y1
output layers.The output of each node propagates from

the input to the outside side. Nonlinear activation I2 V21 w13 w21
a2 c2
functions in multiple layers of neurons allows neural b2
V23 V22 w22 Y1
network to learn nonlinear and linear relationships
between input and output vectors. Each input has an
I3 V31 V32 w31 w32 w23
appropriate weighting W.The sum of W and the bias B
al bm cn
form the input to the transfer function. Any differentiable V33 w33 Y3
activation function f may be used to generate the outputs.
Most commonly used activation function are
purelin f(x)=x, log-sigmoid f(x)=(1+e-x)-1 ,
and tan-sigmoid Figure 2: Three-layered artificial neural system
f(x)=tan(x/2)=(1-e-x) / (1+e-x) When the network has a group of inputs, the updating of
activation values propagates forward from the input
Hyperbolic tangent(Tan-sigmoid) and logistic(log- neurons, through the hidden layer of neurons to the
sigmoid) functions approximate the signum and step output neurons that provide the network response. The
functions,respectively,and yet provide smooth, nonzero outputs can be mathematically represented by:
derivatives with respect to the input signals. These two M N
activation function called sigmoid functions because there Yp = f( ((fXnWnm))*Kmp))
S-shaped curves exhibit smoothness and asymptotic m=-1 n=-1
properties. The activation function fh of the hidden units Yp = The pth output of the network
have to be differentiable functions. If fh is linear, one can Xn = The nth input to the network
always collapse the net to a single layer and thus lose the Wnm = The mth weight factor applied to the nth input
universal approximation/mapping capabilities. Each unit to the network
of the output layer is assumed to have the same activation Kmp = The pth weight factor applied to the mth
function. output of the hidden layer
F( ) = Transfer function (i.e., sigmoid, etc.)
3 BACK PROPAGATION LEARNING The ANS becomes a powerful tool that can be used to
solve difficult process control applications
Error correction learning is most commonly used in
Figure 3 depicts the designing procedure of Artificial
neural networks. The technique of back propagation,
Neural Network Controller.
apply error-correction learning to neural network with
hidden layers.It also determine the value of the learning
rate, ..Values for is restricted such that 0<<1.Back 4 LEARNING ALGORITHMS
propagation requires a perception neural network, (no A gradient-basedalgorithms necessary in the
interlayer or recurrent connection). Each layer must feed development of learning algorithms are presented in this
sequentially into the next layer.In this paper only the section. Learning in neural network is known as learning
three-layer, A, B, and C are investigated. Feeding into rule, in which weights of the networks are incrementally
layer a is the input vector I. Thus layer a has L nodes, ai adjusted so as to improve a predefined performance
(i=1to L), one node for each input parameter. Layer B, the measure over time. Learning process is an optimization
hidden layer, has m nodes, bj (j =1 to m).L = m = 3; in process, it is a search in the multidimensional parameter
practice L m. Each layer may have a different number of (weight) space for solution, Which gradually optimizes an
nodes. Layer C, the output layer, has n nodes, ck (k = 1 to objective(cost)function.
n), with one node for each output parameter. The
interconnecting weight between the ith node of layer A
and the jth node of layer B is denoted as vij, and that
between the jth node of layer B and the kth node of layer C
is wjk.Each node has an internal threshold value. For layer
A, the threshold is TAi, for layer B, TBi, and for layer C,
Tck.The Back propagation neural network is shown in
figure 2..
IJSER 2011
73
ISSN 2229-5518
Figure 4: Flow diagram of learning process

Choose ANN methodology
5 LEVENBERG GRADIENT BASED METHODS

Gradient based methods searches for minima by
Initialize network
parameters(, , no. of neurons),
comparing values of the objective function E( ) at
input and output values different points.The function is evaluated at points
around the current search point and then look for lower
values. Objective function E( ) is minimized in the
adjustable space = [ 1 2, n] and to find minimum
point = *.A given function E depends on an adjustable
Select weight between layers parameter with a nonlinear objective function.
E( )=f( 1 2, n) is so complex that an iterative
algorithm is used to search the adjustable parameter
space efficiently. The next point new is determined by a
Calculate network output
step down from the current point now in the direction
vector d given as:
next= now+d
Calculate MSE (error between actual

Where is some positive step size commonly referred to
output and network output)
as the learning rate.
The principal difference between various descent
algorithms lie in the first procedure for determining
If YES successive directions. After decision is reached, all
algorithms call for movement to a minimum point on the
line determined by the current point now and the
Is
N etwork is trained direction d.For the second procedure, the optimum step
MSE<THRSH
size can be determined by linear minimization as:
If NO
*=arg(min(()))
>0
Change the weights
where
()= E( new +d).
Figure 3: Flow chart for general neural network
algorithm
6 LEVENBERG-MARQUARDT METHOD
The Levenberg_Marquardt algorithm can handle ill-
TRAINING DATA
conditioned matrices well, like nonquadratic objective
functions. Also, if the Hessian matrix is not positive
Target definite, the Newton direction may point towards a local
Error maximum, or a saddle point. The Hessian can be changed
by adding a positive definite matrix I to H in order to
NETWORK + make H positive definite.
IN
Thus,
OUT
next = now (H + I)-1g ,
COST
Weight changes
where I is the identity matrix and H is the Hessian matrix
which is given in terms of Jacobian matrix J as H = JT J.
TRAINING ALGORITHM
Levenberg-marquardt is the modification of the Gauss-
(OPTIMIZATION METHOD)
Newton algorithm as
next = now (JT J)-1 JTr = now (JT J + I)-1 JT r.
The Levenberg-Marquardt algorithm performs initially
small, but robust steps along the steepest descent
direction, and switches to more efficient quadratic Gauss-
IJSER 2011
74
ISSN 2229-5518
Newton steps as the minimum is approached. This networks, all delta values are first evaluated at the output
method combines the speed of Gauss-Newton with the nodes based on the current pattern errors, the hidden
everywhere convergence of gradient descent, and appears values is then evaluated based on the output delta values,
to be fastest for training moderate-sized feedforward and so on backwards to the input layer. Having obtained
neural networks. the node deltas, it is an easy step to find the partial
derivatives dEp/dWij with respect to the weights. The
second factor is dak/dwij because ak is a linear sum, this is
zero if k = i; otherwise
7 FORWARD-PROPAGATION AND BACK- dai
PROPAGATION = Xj
During training, a forward pass takes place.The network dwij
computes an output based on its current inputs.Each The derivative of pattern error Ep with respect to weight
node i computes a weighted ai of its inputs and passes wij is then
this through a nonlinearity to obtain the node input yi dEp
.The error between actual and desired network outputs is = ixj
given by dwij
First the derivative of the network training error with
E = 1 (dpi ypi)2 respect to the weights are calculated. Then a training
2 p i algorithm is performed. This procedure is called back-
where p indexes the pattern in the training set, I indexes propagation since the error signals are obtained
the output nodes, and dpi and ypi are, respectively, the sequentially from the output layer back to the input layer.
desired target and actual network output for the error
with respect to the weights is the sum of the individual
pattern errors and is given as
8 WEIGHT UPDATE ALGORITHM
The reason for updating the weights is to decrease the
dE = dEp = dEp dak error. The weight update relationship is
dWij p dWij p,k dak dWij
where the index k represent all outputs nodes. It is wij = dEp = (dpi ypi) fixj
convenient to first calculate a value i for each node i as
i = dEp = dEp dyk dwij
dai k dyk dai where the learning rate >0 is a small positive constant.
Sometimes is also called the step size parameter.
which measures the contribution of ai to the error on the
current pattern. For simplicity, pattern index p are The Delta Rule is weight update algorithm in the training
omitted on yi, ai and other variables in the subsequent of neural networks. The algorithm progresses
equations. sequentially layer by layer, updating weights as it goes.
For output nodes, dEp / dak, is obtained directly as The update equation is provided by the gradient descent
= -(dpk - ypk) f (for output node). method as
The first term in this equation is obtained from error
equation, and the second term which is wij = wij (k+1)- wij(k) = - dEp
dyk = f (ak) = fk dwij
dak dEp = -(dij ypi)dypi
is just the slope of the node nonlinearity as its current dwij dwij
value. For hidden nodes, i is obtain indirectly as for linear output unit, where
i = dEp = dEp dak = k dak ypi = wij xi
dai k dak dai p dai i
and
where the second factor is obtained by noting that if the dypi = xi
node I connects directly to node k then dak / dai = fiwki,
otherwise it is zero. Thus, dwij
i = fi wk k
k so,
for hidden nodes.i is a weighted sum of the k values of wij = wij (k+1)-wij (k) = (dpi ypi)xi
nodes k to which it has connections wki. The way the The adaptation of those weights which connects the input
nodes are indexed, all delta values can be updated units and the i th output unit is determined by the
through the nodes in the reverse order. In layered corresponding error ei = 1 (dpi ypi)2.
IJSER 2011
75
ISSN 2229-5518
Training Data Analysis

Two training data analysis methods are:
(1)normalizing training set and initializing weights
(2)Principal components analysis(speed up the learning
process).
9 CONCLUSION
The fundamentals of neural network based control
system design are developed in this paper and are
applied to intelligent control of the advanced process.
Intelligent control can also be used for fast and complex
process control problems.
REFERENCES
[1] Rajesh Kumar, Application of artificial neural network in paper
industry, A Ph.D thesis, I.I.T.Roorkee, 2009.
[2] S.I.Amari,N.Murata,K.R.Mullar,M.Fincke,H.H.Yang(1997),Asyptotic
Statistical Theory of overtraining and cross validation,IEEE
Trans.Neural Networks,8(5),985-993.
[3] C.H.Dagli,M.Akay,O.Ersoy,B.R.Fernandez,A.Smith(1997),Intelligent
Engineering Systems Through Artificial Neural Networks,vol.7 of
Neural Networks fuzzy logic Data mining evolutionary Programming.
[4] H.Demuth and M.Beale(1997), Neural Networks toolbox user
guide,mathworks.
[5] L.Fu, Neural Networks in Computer Intelligence(1994).
[6] M.T.Hang.andM.B.Menhaj(1994),Training feedforward Networks with
the Marquardt Algorithm,IEEE Trans. Neural Networks,5(6),989-993.
[7] Hong HelenaMu,Y.P.Kakad,B.G.Sherlock,Application of artificial
Neural Networks in the design of control systems.
IJSER 2011
76
ISSN 2229-5518
Handwritten Character Recognition Using

Neural Network
Chirag I Patel, Ripal Patel, Palak Patel
Abstract Objective is this paper is recognize the characters in a given scanned documents and study the effects of changing
the Models of ANN. Today Neural Networks are mostly used for Pattern Recognition task. The paper describes the behaviors of
different Models of Neural Network used in OCR. OCR is widespread use of Neural Network. We have considered parameters
like number of Hidden Layer, size of Hidden Layer and epochs. We have used Multilayer Feed Forward network with Back
propagation. In Preprocessing we have applied some basic algorithms for segmentation of characters, normalizing of characters
and De-skewing. We have used different Models of Neural Network and applied the test set on each to find the accuracy of the
respective Neural Network.
Index Terms Optical Character Recognition, Artificial Nueral Network, Backpropogation Network, Skew Detection.
1 INTRODUCTION
S uch softwares are useful when we want to convert

our Hard copies into soft copies. Such softwares re-
duces almost 80% of the conversion work while still
some verification is always required.
Optical character recognition, usually abbreviated to
OCR, involves computer software designed to translate
images of typewritten text (usually captured by a scan-
ner) into machine-editable text, or to translate pictures of
characters into a standard encoding scheme representing
them in (ASCII or Unicode). OCR began as a field of re-
search in artificial intelligence and machine vision.
Though academic research in the field continues, the fo-
cus on OCR has shifted to implementation of proven
techniques [4].
2 ARTIFICIAL NUERAL NETWORK

Pattern recognition is extremely difficult to automate. Fig. 1 A simple Neuron
Animals recognize various objects and make sense out of
large amount of visual information, apparently requiring Different choices for weight results in different functions
very little effort. Simulating the task performed by ani- are being evaluated by the network. If in a given network
mals to recognize to the extent allowed by physical limi- whose weight are initial random and given that we know
tations will be enormously profitable for the system. This the task to be accomplished by the network , a learning
necessitates study and simulation of Artificial Neural algorithm must be used to determine the values of the
Network. In Neural Network, each node perform some weight that will achieve the desired task. Learning Algo-
simple computation and each connection conveys a signal rithm qualifies the computing system to be called Artifi-
from one node to another labeled by a number called the cial Neural Network. The node function was predeter-
connection strength or weight indicating the extent to mined to apply specific function on inputs imposing a
which signal is amplified or diminished by the connec- fundamental limitation on the capabilities of the network.
tion. Typical pattern recognition systems are designed using
two pass. The first pass is a feature extractor that finds

features within the data which are specific to the task be-
Chirag I Patel has been completed his M.Tech in Computer Science engi-
neering in Nirma Institute of Technology, Ahmedabad, India, PH-91- ing solved (e.g. finding bars of pixels within an image for
9979541227. E-mail: [email protected] character recognition). The second pass is the classifier,
Ripal Patel has been completed her M.E in Electronics & C ommunication which is more general purpose and can be trained using a
engineering in Dharmsinh Desai Institute of Technology, Nadiad, India,
PH-91-9998389428.E-mail: [email protected]
neural network and sample data sets. Clearly, the feature
Palak Patel is pursuing her M.E in Electronics & C ommunication engi- extractor typically requires the most design effort, since it
neering in G.H.Patel College of Enginnering & Technology, Vallabh Vi- usually must be hand-crafted based on what the applica-
dyanagar, India, PH-91-9998389428.E-mail: [email protected]
IJSER 2011
77
ISSN 2229-5518
tion is trying to achieve. Firstly we needed a method to extract a given character

from the document. For this purpose we modified the
One of the main contributions of neural networks to pat- graphics 8-way connected algorithm (which we call as
tern recognition has been to provide an alternative to this EdgeDetection).
design: properly designed multi-layer networks can learn
complex mappings in high-dimensional spaces without
requiring complicated hand-crafted feature extractors.
Thus, rather than building complex feature detection al-
5 PREPROCESSING
gorithms, this paper focuses on implementing a standard 5.1 Character Extraction Algorithm
backpropagation neural network. It also encapsulates the
Preprocessing that is required for effective. 1. Create a TraverseList :- List of pixels which have
been already traversed. This list is initially emp-
2.1 Backpropogation ty.
Backpropagation was created by generalizing the Wi- 2. Scan row Pixel-by-Pixel.
drow-Hoff learning rule to multiple-layer networks and 3. Whenever we get a black pixel check whether the
nonlinear differentiable transfer functions. Input vectors pixel is already in the traverse list, if it is simply
and the corresponding target vectors are used to train a ignore and move on else apply Edgedetection
network until it can approximate a function, associate Algorithm.
input vectors with specific output vectors, or classify in- 4. Add the List of Pixels returned by Edgedetection
put vectors in an appropriate way as defined by you. Algorithm to TraverseList.
Networks with biases, a sigmoid layer, and a linear out- 5. Continue the steps 2 - 5 for all rows
put layer are capable of approximating any function with
a finite number of discontinuities. 5.2 Edge Detection Algorithm
The Edge Detection Algorithm has a list called traverse
list. It is the list of pixel already traversed by the algo-
3 ANALYSIS rithm.
By analyzing the OCR we have found some parameter
which affects the accuracy of OCR system [1][5]. The pa- EdgeDetection(x,y,TraverseList);
rameters listed in these papers are skewing, slanting, 1) Add the current pixel to TraverseList. The cur-
thickening, cursive handwriting, joint characters. If all rent position of pixel is (x,y).
these parameters are taken care in the preprocessing 2) NewTraverseList= TraverseList + current posi-
phase then overall accuracy of the Neural Network tion (x,y).
would increase.
If pixel at (x-1,y-1) then
Check if it is not in TraverseList.
4 DESIGN AND IMPLEMENTATION Edgedetection(x-1,y-1,NewTraverseList);
Initially we are making the Algorithm of Character Ex- endif
traction. We are using MATLAB as tool for implementing
the algorithm. Then we design neural network, we need If pixel at (x-1,y) then
to have a Neural Network that would give the optimum Check if it is not in TraverseList.
results [2]. There is no specific way of finding the correct Edgedetection(x-1,y,NewTraverseList);
model of Neural Network. It could only be found by trial endif
and error method. Take different models of Neural Net-
work, train it and note the output accuracy. If pixel at (x-1,y) then
There are basically two main phases in our Paper: Check if it is not in TraverseList.
Preprocessing and Character Recognition . Edgedetection(x-1,y+1,NewTraverseList);
endif
In first phase we have are preprocessing the given
scanned document for separating the Characters from it If pixel at (x,y-1) then
and normalizing each characters. Initially we specify an Check if it is not in TraverseList.
input image file, which is opened for reading and prepro- Edgedetection(x,y-1,NewTraverseList);
cessing. The image would be in RGB format (usually) so Endif
we convert it into binary format. To do this, it converts
the input image to grayscale format (if it is not already an If pixel at (x,y+1) then
intensity image), and then uses threshold to convert this Check if it is not in TraverseList.
grayscale image to binary i.e all the pixels above certain Edgedetection(x,y+1,NewTraverseList);
threshold as 1 and below it as 0. endif
IJSER 2011
78
ISSN 2229-5518
tered before i.e. it should not be in the TraverseList.

If pixel at (x+1,y-1) then 5.3 Normalizing
Check if it is not in TraverseList. Now as we have extracted the character we need to
Edgedetection(x+1,y-1,NewTraverseList); normalize the size of the characters. There are large varia-
endif tions in the sizes of each Character hence we need a me-
thod to normalize the size.
If pixel at (x+1,y) then We have found a simple method to implement the nor-
Check if it is not in TraverseList. malizing. To understand this method considers an exam-
Edgedetection(x+1,y,NewTraverseList); ple that we have extracted a character of size 7 X 8. We
endif want to convert it to size of 10 X 10. So we make a matrix
of 70 X 80 by duplicating rows and columns. Now we
If pixel at (x+1,y+1) then divide this 70 X 80 into sub Matrix of 7 X 8. We extract
Check if it is not in TraverseList. each sub matrix and calculate the no. of ones in that sub
Edgedetection(x+1,y+1,NewTraverseList); matrix. If the no. of ones is greater than half the size of
endif sub matrix we assign 1 to corresponding position in nor-
malized matrix. Hence the output would be a 10 X 10 ma-
3) return; trix.
The EdgeDetection algorithm terminates when it has

covered all the pixels of the character as every pixels po-
sition would be in TraverseList so any further call to Ed-
geDetection is prevented.
Fig 5(a) shows original representation of the character.
Fig 4(a) shows the traversing each scan lines.
Fig 5(b) shows the Normalized Character representation

after Normalizing.
The Fig 5(a) is shows a representation of character of 12 X
12 size. Using the above algorithm it is converted into a
character of 8 X 8 as shown in the Fig 5(b).
5.4 Skew Detection

The Characters are often found to be skewed. This
would impose problems on the efficient character recog-
4(b) shows the respective calls made to the all 8- nition [3]. So to correct the effect of this skewedness we
neighbouring pixels. need counter rotate the image by an angle .
The Edge detection is called when we hit a pixel (i.e. en- We use a very simple but effective technique for Skew
counter a pixel with value 1). As per the algorithm the Correction. We use Line Fitting i.e. Linear Regression to
current position is entered in TraverseList and recursive find the angle . Consider the Skewed character as a
calls are made to all 8 - neighboring pixels. Before the graph i.e. all the pixels that have value 1 are considered to
calls are made it is ensured that the corresponding neigh- be data points. Then we perform linear regression using
boring pixels is having value 1 and is not already encoun- the equation Y = M*X +C. Using the formulas for regres-
IJSER 2011
79
ISSN 2229-5518
sion we calculate M= (nxiyi - xiyi) / (nxi2-(xi)2). are 26. The no. of nodes of input layer are 100 and the no.
This angle is equivalent to the skewed angle so by rotat- of node of output layer are 26. The no. of hidden layer and
ing the image by opposite of this angle will remove the the size of hidden layer vary.
skew ness. This is a very crude way of removing skew Fig 7. The general Model of ANN used.
ness there are other highly efficient ways of removing
skew ness. But for Characters that have very low Skew
angles this gets the thing done.
The Characters are often found to be skewed. This

would impose problems on the efficient character recog-
nition. So to correct the effect of this skewed ness we
need counter rotate the image by an angle .
.We use a very simple but effective technique for Skew

Correction. We use Line Fitting i.e. Linear Regression to
find the angle . Consider the Skewed character as a
graph i.e. all the pixels that have value 1 are considered to
be data points. Then we perform linear regression using
the equation Y = M*X +C. Using the formulas for regres-
Fig 8.
sion we calculate M= (nxiyi - xiyi) / (nxi2-(xi)2).
Ttrai
This angle is equivalent to the skewed angle so by rotat-
ning
ing the image by opposite of this angle will remove the
au-
skew ness. This is a very crude way of removing skew
tomated character extraction
ness there are other highly efficient ways of removing
skew ness. But for Characters that have very low Skew
angles this gets the thing done.
Fig
6(a)
Skew
ed
Im-
age Fig 6(b) Corrected Image.
6 NUERAL NETWORK DESIGN

For training and simulating purposes we have scanned
certain documents. We have 2 types of documents train Fig 9 recognition
documents and test documents. The train documents are
the images of the documents which we want to use for
training. Similarly test documents are the images of docu-
ments which we want to use for test. According to the cha-
racters in the documents we train the neural network and
apply the test documents.
We have different Models of Neural Network. Hence we
record certain parameters like training time, accuracy etc.
to find the effectiveness of the Neural Network.
We have selected an image size of 10 X 10 as an input to the
Neural Network. Hence we have taken a neural network
that has 100 inputs. We are performing the test on only Fig 10 Training User defined Character Extraction.
Capital characters so the outputs of the Neural Networks
7 TEST AND RESULTS ANALYSIS
7.1 Test
This section shows some implementation results. The
IJSER 2011
80
ISSN 2229-5518
training variables involved in the tests were: the number

of cycles, the size of the hidden layer, and the number of
Epochs Number of Configuration Accuracy
hidden layer. The dataset consisted of A-Z typed charac-
Hidden (No. of nodes in (%)
ters of different size and type. Thus the input layer con-
Layer HL)
sisted of 100 neurons, and the output layer 26 neurons
(one for each character). Ideally, wed like our training 300 3 26-52-26 31
and testing data to consist of thousands of samples, but
this not feasible since this data was created from scratch. 600 3 26-52-26 65
1000 3 26-52-26 82
Table 1. Model 1
Epochs 300 3 26-52-78 29
Number of Configuration Accuracy
600 3 26-52-78 74
Hidden Layer (No. of nodes in HL) (%)
1000 3 26-52-78 92
300 1 26 20 300 3 78-26-78 27
600 1 26 65 600 3 78-26-78 71
1000 1 26 82 1000 3 78-26-78 91
300 1 52 25 Table 4. Model 4
600 1 52 69
1000 1 52 88 Epochs Number of Configuration
Accuracy
Hidden (No. of nodes (%)
300 1 78 27 Layer in HL)
600 1 78 71 300 3 26-52-26 31
1000 1 78 91 600 3 26-52-26 65
1000 3 26-52-26 82
Table 2. Model 2
300 3 26-52-78 29
600 3 26-52-78 74
1000 3 26-52-78 92
Epochs 300 3 78-26-78 27
Number of Configuration Accuracy
Hidden Layer (No. of nodes in HL) (%) 600 3 78-26-78 71
1000 3 78-26-78 91
300 2 26-52 23
600 2 26-52 67
1000 2 26-52 81
300 2 52-78 40
600 2 52-78 78
1000 2 52-78 96
300 2 26-78 27
600 2 26-78 77
1000 2 26-78 89
Table 5. Model 5
Table 3. Model 3
IJSER 2011
81
ISSN 2229-5518
A small number of nodes in the hidden layer (eg. 26)

Epochs Number of Configuration
Accuracy lower the accuracy.
Hidden (No. of nodes (%) A large number of neurons in the hidden layer help in
Layer in HL) increasing the accuracy; however there is probably some
upper limit to this which is dependent on the data being
300 4 26-52-78-104 35
used. Additionally, high neuron counts in the hidden
600 4 26-52-78-104 79 layers increase training time significantly.
As number of hidden layer increases the accuracy in-
1000 4 26-52-78-104 96 creases initially and then saturates at certain rate proba-
300 4 26-52-78-26 30 bly due to the data used in training.
Mostly Accuracy is increased by increasing the number
600 4 26-52-78-26 61 of cycles.
1000 4 26-52-78-26 86 Accuracy could also be increased by increasing the
training set.
300 4 78-26-52-104 43
600 4 78-26-52-104 82 7.3 Additional Formatting and Style Resources
1000 4 78-26-52-104 98 Additional information on formatting and style issues can be
obtained in the IJSER Style Guide, which is posted online at:
300 4 78-26-78-52 31 http://www.ijser.org/. Click on the appropriate topic under the
600 4 78-26-78-52 88 Special Sections link.
1000 4 78-26-78-52 94
8 CONCLUSION
The backpropagation neural network discussed and im-
Table 6. Model 6
plemented in this paper can also be used for almost any
general image recognition applications such as face detec-
Epochs Number of Configuration
Accuracy
tion and fingerprint detection. The implementation of the
Hidden (No. of nodes (%)
fully connected backpropagation network gave reasona-
Layer in HL)
ble results toward recognizing characters.
300 4 26-52-78-104 35
The most notable is the fact that it cannot handle major
600 4 26-52-78-104 79
variations in translation, rotation, or scale. While a few
1000 4 26-52-78-104 96 pre-processing steps can be implemented in order to ac-
count for these variances, as we did. In general they are
300 4 26-52-78-26 30 difficult to solve completely.
600 4 26-52-78-26 61
REFERENCES
1000 4 26-52-78-26 86
300 4 78-26-52-104 43 [1] S. Basavaraj Patil, N. V. Subbareddy Neural network based
system for script identification in Indian documents in Sadha-
600 4 78-26-52-104 82
na Vol. 27, Part 1, February 2002, pp. 8397.
1000 4 78-26-52-104 98 [2] T. V. Ashwin, P. S. Sastry A font and size-independent OCR
system for printed Kannada documents using support vector
300 4 78-26-78-52 31
machines in Sadhana Vol. 27, Part 1, February 2002, pp. 3558.
600 4 78-26-78-52 88 [3] Kavallieratou, E.; Fakotakis, N.; Kokkinakis, G., New algo-
rithms for skewing correction and slant removal on word-level
1000 4 78-26-78-52 94
[OCR] in Proceedings of ICECS '99.
[4] Simmon Tanner, Deciding whether Optical Character Recogni-
We have used sigmoid transfer function in all the layers. tion is Feasible.
We have used same dataset for training all the different [5] Matthew Ziegler, Handwritten Numeral Recognition via
Models while testing character set was changed. Neural Networks with Novel Preprocessing Schemes.
7.2 Result Analysis

From the results, the following observations are made:
IJSER 2011
82
International Journal of Scientific & Engineering Research, Volume 2, Issue 4, April-2011
ISSN 2229-5518
Improved Performance of M-ary PPM in Different

Free-Space Optical Channels due to Reed Solomon
Code Using APD
Nazmi A. Mohammed, Mohammed R. Abaza and Moustafa H. Aly
Abstract Atmospheric turbulence induced fading is one of the main impairments affecting the operation of free-space optical
(FSO) communication systems. In this paper, the bit error rate (BER) of M-ary pulse position modulation (M-ary PPM) of direct-
detection and avalanche photodiode (APD) based is analyzed. Both log-normal and negative exponential fading channels are
evaluated. The investigation discusses how the BER performance is affected by the atmospheric conditions and other
parameters such as the forward error correction using Reed Solomon (RS) codes and increasing Modulation level. Results
strongly indicate that, RS-coded M-ary PPM are well performing for the FSO links as it reduces the average power required per
-9
bit to achieve a BER below 10 in both turbulence channels.
Index Terms Free Space Optics (FSO), M-ary Pulse Position Modulation (M-ary PPM), Reed Solomon (RS) codes, Log
Normal Channel, Negative Exponential Channel, Avalanche Photodiode (APD).
1 INTRODUCTION
F ree space optical (FSO) systems have been widely

deployed for inter-satellite and deep-space communi-
cations. In recent years, however, because of its nu-
for FSO is the on-off keying (OOK) which offers band-
width efficiency but lacks power efficiency. Binary level
signaling though is the simplest and most common mod-
merous advantages over radio-frequency (RF) technology ulation scheme for the optical intensity channel and offers
such as extremely high bandwidth, license-free and inter- low power efficiency and high bandwidth efficiency.
ference immunity, FSO has attracted considerable atten- Power efficiency as well as the improved system perfor-
tion for a variety of applications, e.g. last mile connectivi- mance can be achieved by adopting pulse position mod-
ty, optical-fiber backup and enterprise connectivity. In ulation (PPM) schemes. M-ary PPM achieves high power
such kind of applications, FSO systems basically utilize efficiency at the expense of reduced bandwidth efficiency
atmosphere as transmission medium rather than the free compared with other modulation schemes. The optimal
space. So, the performance of FSO link is inherently af- PPM order is high, since a higher order modulation
fected by atmospheric conditions. Among these condi- creates the higher peak power needed to overcome the
tions, atmospheric turbulence has the most significant weak average power. M-ary PPM has been previously
effect. It causes random fluctuations at the received signal suggested as a suitable modulation scheme for FSO sys-
intensity, i.e., channel fading, which leads to an increase tems [2]. The IrDA specification for the 4 Mbps short dis-
in the bit error rate (BER) of the optical link [1]. tance wireless infrared links specifies a 4-PPM modula-
tion scheme [3].
Current FSO communication systems employ intensi-
ty modulation with direct detection (IM/DD) and use Reed Solomon (RS) codes are a class of block codes
light emitting diodes (LED) or laser diodes as transmitters that operate on symbols rather than bits. So, RS codes can
and PIN photodiode or avalanche photodetectors (APD) correct both random bit errors and burst symbol errors.
as receivers. These devices modulate and detect solely the Moreover, their hard decoding algorithm can be easily
intensity of the carrier and not its phase. Furthermore, implemented even at a high operation speed. Internation-
biological safety reasons constrain the average radiated al Telecommunication Union-Telecommunication (ITU-T)
optical power, thereby constraining the average signal standard forward error correction (FEC) scheme based on
amplitude. The most reported modulation technique used RS (255,239) codes has been widely used in l0 Gbps prac-
tical optical fiber transmission systems [4]. Kiasaleh de-
rived upper bounds on the BER of M-ary PPM over log-
Nazmi A. Mohammed is with the Electronics and Communications Engi- normal and exponential distributed channels, when an
neering Department, Arab Academy for Science, Technology and Maritime APD is used [5].
Transport, Egypt. [email protected].
Mohammed R. Abaza is with the Electronics and Communications Engi- In this paper, Kiasaleh expressions are used to com-
neering Department, Arab Academy for Science, Technology and Maritime pare the effect of the BER of average photons per PPM bit
Transport, Egypt. OSA Student Member. [email protected]. with the former channels without coding and with RS
Mostafa H. Aly is with the Electronics and Communications Engineering
(255,207) coding. The remainder of the paper is organized
Department, Arab Academy for Science, Technology and Maritime Trans-
port, Egypt. OSA Member. [email protected]. as follows. The models of FSO channels are presented in
Section 2. Based on the theory presented, a numerical
IJSER 2011
83
ISSN 2229-5518
analysis of the M-ary PPM is carried out in Section 3. This The variance, 2n, of the thermal noise in a PPM slot is
is followed by the main conclusions in Section 4. defined by [5]
2 K T Tslot
2 MODELS OF FSO CHANNELS 2n (6)
RL
2.1 Log-Normal Channel
The log-normal channel is classified as weak turbu- where T is the effective absolute temperature of the rceiv-
lence, which is characterized by a scintillation index less er, K is Boltzmann constant, RL is the APD load resistance
than 0.75. In general, the scintillation index is a compli- and Tslot is the PPM slot duration which is related to the
cated function of the beam parameters, propagation dis- data rate, Rb, by [5]
tance, heights of the transmitter and receiver, and the
fluctuations in the index of refraction. In fact, the main log 2 M
Tslot (7)
source of scintillation is due to fluctuations (due to tem- M Rb
perature variations) in the index of refraction, which is
commonly known as optical turbulence. The log-normal In case of coding, Rb must be multiplied by (n/k), where
model is also valid for propagation distances less than 100 n is the codeword length and k is the message length.
m [5].
The bit error rate (BER) of an M-ary PPM in log-normal The symbol error rate (Psymbol) can be calculated from bit
channel is given by [6] error rate (Pb) as [8]
N
M e 2( 2 k x i m k ) 2 M 1
PbM
2 i N, i 0
w i Q
Fe 2 k x i m k K n
(1)

Psymbol Pb
M
(8)

The probability of the uncorrectable symbol error (Pues)

where M is the modulation level, wi and xi are the weight due to RS codes can be calculated by the formula [9, 10]
factors and the zeros of the Hermite polynomial [7].
1 n n i
The scintillation index (2SI) as a function of the variance
i Pq 1 Pq
ni
of the log-normal channel (2k) is given by [5]
Pues
n i t 1 k
(9)
2
2
SI e k 1 (2) where t =((n-k)/2) is the symbol error correcting capabili-
ty, Pq is the q-bit RS symbol error probability.
The average photons per PPM slot (E{Ks}) are functions of
the mean (mk) and the variance of the log-normal channel The BER after coding (Pbc) is given by [9]
and have the form [5].
n 1
2k

mk
Pbc Pues (10)
EK s e
2

(3) 2 n
The total noise photons per PPM slot, Kn, which results 2.2 Negative Exponential Channel
from background noise and thermal noise, is [5] The negative exponential channel is classified as strong
turbulence, which is characterized by a scintillation index
2 2n greater than 1. The negative exponential model is valid for
Kn 2FK b (4) propagation distances more than 100 m or several kilometers
Egq 2 [5, 8].
The BER of the negative exponential channel, PbM, is giv-
where Kb is the average background noise photons per
en by [5].
PPM slot, E{g}is the average gain of the APD and q is the
electron charge.
M N EK s x i2
The noise factor, F, of the APD is defined by [5] PbM i i
w x Q (11

F EK s x i K n
2 i N,i 0 2

F 2 Eg (5) )
where is the ionization factor. To get the BER after coding (Pbc) due to negative exponen-
tial channel using RS (255,207), we apply the same proce-
dure of the log-normal channel coding steps, which are
IJSER 2011
84
ISSN 2229-5518
mentioned in (7), (8), (9), (10) and use this in (11). At a BER of 10-9, the value of average photons per PPM
bit in strong turbulence is found 728 without coding and
141 with coding giving an improvement of 11.61 dB.
3 NUMERICAL RESULTS AND DISCUSSIONS While the value of average photons per PPM bit in weak
turbulence is found 543 without coding and 32 with cod-
Based on the described model, the BER of 10-9, which ing which gives an improvement of 16.6 dB. It is shown
is considered as a practical performance target for FSO that, coded strong turbulence 8-PPM outperforms weak
link [11], is calculated for log-normal channel and nega- turbulence 8-PPM without coding at BER less than 10-4.
tive exponential channel and the obtained results are dis-
played in Figs. 1-3. In these figures, the value of scintilla-
tion index (2SI) is taken 0.3 for weak turbulence and 1 for
strong turbulence. The values of other parameters are
taken as: BER=2.4 Gbps, Kb =10 photons per PPM slot,
RL=50 , =0.028, T=300K, E{g}=150, n=255, k=207, t=24
symbols.
Variations of BER with the average number of pho-
tons received per PPM bit (logarithmic), which are equal
log10(E{Ks}/number of bits in PPM symbol), are shown in
the following figures. In discussions, all results of average
photons per PPM bit are in numerical values.
In Fig. 1, binary PPM (BPPM) is used to compare be-
tween the effect of weak turbulence and strong turbu-
lence without coding and with RS (255,207) coding.
Fig. 2. Comparison of non-coded 8-PPM in weak turbulence of 0.3

scintillation index with Reed Solomon (255,207) and in strong turbu-
lence of 1.0 scintillation index.
In Fig. 3, 256-PPM is used for weak turbulence and

strong turbulence without coding and with RS (255,207)
coding.
Fig.1. Comparison of non-coded BPPM in weak turbulence of 0.3

scintillation index with Reed Solomon (255,207) and in strong turbu-
lence of 1.0 scintillation index.
At a BER of 10-9, the value of average photons per PPM

bit in strong turbulence is found 2046 without coding and
219 with coding which gives an improvement of 9.7 dB. Fig. 3. Comparison of non-coded 256-PPM in weak turbulence
of 0.3 scintillation index with Reed Solomon (255,207) and in
While the value of average photons per PPM bit in weak strong turbulence of 1.0 scintillation index.
turbulence is found 1462 without coding and 52 with cod-
ing which gives an improvement of 14.49 dB. It also At a BER of 10-9, the value of average photons per PPM
shows that, coded strong turbulence BPPM outperforms bit in strong turbulence is found 308 without coding and
weak turbulence BPPM without coding at a BER less than 80 with coding showing an improvement of 14.08 dB.
10-3. While, in weak turbulence, the value of average photons
per PPM bit is found 258 without coding and 25 with cod-
In Fig. 2, 8-PPM is used to show the effect of multile- ing giving an improvement of 17.67 dB. It also shows that,
vel modulation and to compare between the effect of coded strong turbulence 256-PPM outperforms weak tur-
weak turbulence and strong turbulence without coding bulence 256-PPM coding less at BER less than 10-4.
and with RS (255,207) coding.
IJSER 2011
85
ISSN 2229-5518
The obtained results for coded and non-coded M-ary ACKNOWLEDGMENT

PPM are summarized and compared in Tables 1 and 2 for
Authors wish to thank Prof. Kamran Kiasaleh from Uni-
strong and weak turbulence, respectively.
versity of Texas at Dallas, USA, for his fruitful discus-
TABLE 1 sions.
Average number of photons per ppm bit required achiev-
ing ber 10-9 in strong turbulence. REFERENCES
[1] Yi Xiang, Liu. Zengji, Yue Peng and Shang Tao, BER performance
M-ary PPM Strong Strong turbulence Improvement
turbulence with RS due to RS code + analysis for M-ary PPM over gamma-gamma atmospheric turbulence
code multilevel modulation channels, Proceeding of the 6th International Conference on Net-
in Strong turbulence working and Mobile Computing (WiCom), Chengdu, pp. 1-4, 2010.
respect to non coded [2] S. Sheikh Muhammad, Tomaz Javornik, Igor Jelovcan, Zabih Ghas-
BPPM semlooy and Erich Leitgeb, Comparison of hard-decision and soft-
% dB decision channel coded M-ary PPM performance over free space opti-
cal links, Eur. Trans. Telecomms., vol. 20, pp. 746757, 2008.
BPPM 2046 219 89.3 9.7 [3] S. Hranilovic, Wireless Optical Communication Systems, Springer
8-PPM 728 141 93 11.61 Science + Business Media, Boston, 2005.
[4] Zheng Zheng Juanjuan Yan and Anshi Xu Weiwei Hu, Improved
256-PPM 308 80 96 14.08 performance of M-ary PPM free-space optical communication systems
in atmospheric turbulence due to forward error correction, Proceed-
ing of the 10th International Conference on Communication Technolo-
TABLE 2 gy (ICCT), Guilin, pp.1- 4, 2006.
Average number of photons per ppm bit required achiev- [5] Kamran Kiasaleh, Performance of APD-based, PPM free-space opti-
ing ber 10-9 in weak turbulence. cal communication systems in atmospheric turbulence, IEEE Trans-
actions on Communications, vol. 53, pp.1455-1461, Sep. 2005.
M-ary PPM Weak Weak turbulence Improvement [6] Mohammed R. Abaza and Kamran Kiasaleh, Erratum to Perfor-
turbulence with RS due to RS code + mance of APD-based, PPM free-space optical communication systems
code multilevel modulation in atmospheric turbulence, IEEE Transactions on Communications,
in Weak turbulence submitted for publication.
respect to non coded [7] Philip M. Morse, et al., Handbook of Mathematical Functions with.
BPPM
Formulas, Graphs, and Mathematical Tables, vol. 10, Milton Abra-
% dB mowitz and Irene A. Stegun, Eds. Washington, D.C.: U.S. Govern-
BPPM 1462 96.4 14.49 ment Printing Office, p. 924, 1972.
52
[8] Jagtar Singh and V.K.Jain, Performance analysis of BPPM and M-
8-PPM 543 32 97.8 16.6 ary PPM optical communication system in atmospheric turbulence,
256-PPM 258 IETE Tech. Review, vol.25, no.4, pp. 145-152, July-Aug. 2008.
25 98.3 17.67
[9] J. G. Proakis, Digital Communications, McGraw-Hill, New York,
NY, 1983.
[10] T. Aaron Gulliver, Matched Q-ary Reed-Solomon Codes with M-ary
All results obtained indicate that, RS codes have low per- Modulation, IEEE Transactions on Communications, vol. 45,
formance in low photons per bit range. This is because the pp.1349-1353, Nov. 1997.
low photons per bit causes burst erros which are beyond [11] Murat Uysal and Jing Li, BER performance of Coded Free-Space
the correction capability of RS codes. Under these condi- Optical Links over Strong Turbulence Channels, Proceeding of the
tions, spreading the errors by using a code matched inter- 59th Vehicular Technology Conference (VTC), Milan, pp. 352-356,
leaver and using concatenated codes will make the cod- 2004.
[12] S. Sheikh Muhammad, et al., Reed Solomon coded PPM for Terre-
ing more effective [12]. strial FSO Links, Proceeding of the 7th International Conference on
Electrical Engineering (ICEE), Lahore, pp. 1-5, 2007.
4 CONCLUSION
In this paper, the performance of FSO system with M-ary
PPM based on RS codes scheme has been numerically
analyzed in weak and strong atmospheric turbulence. The
results show that, the average number of photons per bit
at 10-9 BER has been improved by 14.08 dB (compared
with the value of BPPM without coding) in strong turbu-
lence and to 17.67 dB in weak turbulence using RS
(255,207) + 256-PPM.
The results also show that, coded strong turbulence out-
performs the non-coded weak turbulence without coding
for BPPM, 8-PPM and 256-PPM at 10-9 BER, which indi-
cates a great improvement of the performance of the sys-
tem tolerance for the intensity fluctuations induced by
atmospheric turbulence. RS codes can be combated with
matched interleaver concatenated coding to solve the
problem in low photons per bit range.
IJSER 2011
86
ISSN 2229-5518
Electricity Sector Restructuring Experience of

Different Countries
Archana Singh, Prof. D.S.Chauhan
Abstract Electricity Market from economic, regulatory and engineering perspective is a very demanding system to control
.There is requirement of provision of cost efficiency, lower impact of environment alongwith maintenance of security of supply for
use of competition and regulation in the electricity market. Many countries due to failure of its system for adequately
management of electricity companies, followed restructuring for its electricity sector. In various countries, different restructuring
models were experimented but in the initial phase restructuring was opposed by the parties favouring existing vertically
integrated electricity sector. In the paper , restructuring experience of different countries are outlined .
Index TermsDeregulation, Wholesale Electricity Market, Forward Markets, Independent system Operator, Power Exchange.
1 INTRODUCTION
sumption) from transmission as considering different
Electric utilities have been vertically integrated mo- functions associated with selling and buying electric
nopolies that have combined generation, transmission energy. The reason behind separation of transmission
and distribution facilities to serve the needs of the cus- which is the means of transporting the tradable com-
tomer in their service territories. The price of electricity modity and ability to influence the transmission- use
through, for example, line ratings, line maintenance
was traditionally set by a regulatory process, rather than
schedules and network data would be to avoid very
using market forces, which were designed to recover the
powerful competitive advantage to a participant. Beside
cost of producing and delivering electricity to customers this, another important function is system operation
as well as the capital cost. Due to this monopolistic ser- which is traditionally viewed as a genera-
vice regime, customers had no choice of supplier; and tion/transmission function. This function has evolved to
suppliers were not free to pursue outside their designat- the Independent System Operator (ISO) in the most elec-
ed service territories. The main reason for deregulation tricity markets presently which is responsible for coor-
in developing countries has been to provide electricity dinating maintenance schedules and performing securi-
to customers at lower prices, and to open the market for ty assessment.
competition by allowing smaller players to have access The deregulation processes have been started with
to the electricity market by reducing the share of large debate for defending the vertically integrated model
state owned utilities. On the other hand ,high growth in from opposition by private and state monopolies [3].
demand and irrational tariff policies have been the driv- The first was Chile to start effort in 1980s for restructur-
ing its electricity sector. The most discussed deregula-
ing forces for the deregulation in developing countries
tion was the British one, with more interest in Norway
.Technical and managerial inefficiencies in these coun-
Model and much attention to actions in Unites States,
tries have made it difficult to sustain generation and especially California State. In South America a major
transmission expansions and hence many utilities were transformation took place throughout the electric power
forced by international funding agencies to restructure industry from 1980 onwards (chronological progress-
their power industries[1]. shown in fig.1).
Eletricity markets are having a very important characte-
ristic of its organizational structure which has been ac-
commodated as the most significant change in the in-
2 RESTRUCTUING EXPERIENCE OF DIFFERENT
dustry. Vertically integrated industry structure (a regu- COUNTRIES
lated monopoly) as the traditional industry structure The electricity sector reform in many developed countries
was owned and operated as a single organization for have already undertaken since the 1980s. Initially it was
distribution, transmission, and generation functions[2]. not clear to how to increase efficiency by electricity sector
However, the vertically integrated structure, by virtue of reform. As a matter of fact over various countries, there
the fact that it is a monopolistic structure, is not amend- exists diversity in the wholesale electricity market opera-
able to introduction of competition. tion. A transparent, open marketplace would encourage
Current industry structure primarily requires separate competition among generators and reveal the inefficien-
functions of the generation and distribution (or con- cies of the current system to improve the efficiency of the
IJSER 2011
87
ISSN 2229-5518
electricity sector. tions between transmission and commercialization bid-

ders are made under two modalities which are through
subscription of guaranteed bilateral contracts and
through direct transcations in the energy stock market
,free offer and demand .Projects of expansion plan are
assigned through a scheme of public bids.There is open
access to the NTS(National Transmssion System)network
[5].
Transmission charges are based on connection charges
and use of network charges. Evolution of the electricity
sector has multifolded by significant increase in private
participation.Competition in commercialization to unre-
gulated users has incremented.
2.3 Argentina Market

Fig.1. Chronological progression of developments of
power markets across the world. Power sector restructuring activities in Argentina
started in 1990, resulting in the enactment of the Electric
2.1 Chile Market
Power Regulatory Frame .It created the National Regula-
tory Agency (ENRE) and wholesale Power Market.
Hydropower production in Chile varies from 60% to Transco and Disco required license to operate. Generation
95%.For supply, there are 4 Gencos with one owned by developed as free activity in competition and having the
State Government. Gencos are dispatched depending on transmission network as open access. Marginal declared
audited generation costs and reservoir levels.There are 3 costs was basis for dispatch on recognition of remunera-
Transcos and private 13 Discos. A poolco which is Eco- tion for capacity as a function of system failure risks.
nomic load dispatch center (CDEC) forecasts global de- Transport is organized as a monopolistic activity with
mand and updates values monthly.The system operator national network (Transener) and six concessionaires for
in charge of coordinating grid operation is run only by regional network(Distros);it included concession contract
generators favoring the development of several private , transport tariffs based on the economic cost of losses
transmitters. In Chile, wholesale market includes a uni- (node factors) and network unavailability (adaptation
fied market place, a transport system that carries power factors), plus network O &M costs; expansion at the ex-
and market prices defined a spot price for each node pense of interested parties ;failure penalties as a function
along grid. A poolco manages dispatch, reliability and of transport charges. It was mandatory for Transport con-
pooling functions .The National Energy Commission cessionaries to provide nondiscriminatory open access to
(CNE) provides arbitration in the case of disputes [4]. their transmission system. All existing and future load
Losses including energy theft were halved in Chile within must to be served by distribution concessionaries in their
just seven years and Number of customers per distribu- concession area[6]. Wholesale electricity market is under-
tion worker has multifolded in ten years after restructur- taken by private company (Cammesa) whose share hold-
ing. ers are the associations of generators, distributors, trans-
porters, the national state with equal distribution of
shares and the large users .
2.2 Colombia Market 2.4 Australia Market

Wholesale Electricity Market came into operation in July
The process of restructuring of the electricity industry in
1995. Rules promoting free competition in the generation
Australia was initiated in 1991, and by 1998 a National
and commercialization business were implemented.
Electricity Market was developed, where the National
Transmission and distribution business were treated as
Electricity Management Company (NEMCO) acted as
monopolies and competitions were implemented whe-
both the ISO and IMO. Generators could sell energy ei-
rever possible. For the reservoirs of the national intercon-
ther by bidding in the spot market, or through formal
nected system (NIS) defining minimum operative levels a
(bilateral) contracts. The most extensive restructuring is
methodology was ruled. CREG (Energy Regulation
occurring in the South Australia, Victoria, New South
Commission Group) defined limits for the horizontal and
Wales and Queensland to form National Electricity Mar-
vertical integration business. An energy stock market
ket (NEM).The key aspect of transition process were like
(generator pool) and a central operator (National Dis-
elimination of barriers to entry and of barriers to trade
patch Centre) of the NIS to support its operation. Transac-
between states, creation of pool style (bulk electricity
IJSER 2011
88
ISSN 2229-5518
market) competitive entities in generation and in retail independent company. Swedish electricity market un-
supply and development of regulatory arrangements ap- bundled in 1996. Thereafter, a common electricity ex-
propriate to the new regime [7]. Table1 below summariz- change for Norway and Sweden was established under
es the deregulated structure in Australia. the name of Nord Pool. In 1998, Finland effectively en-
NSW(New Victoria South Queensland tered into Nordic Market[8]. Denmark joined Nord Pool
South Australia subsequently. Nord Pool is owned by the Transmission
Wales) System operators (TSO) of Norway and Sweden. Nord
Pool provides freedom of choice to the large consumers. It
Genera- NSWs Five ex- ET- Three gene- organizes trade in standardized physical and financial
tion three ex- SECV(Sta SA(Electr rating com- power contracts. Close cooperation between the system
ECNSW(Ele te Elec- icity panies. operation and market operation is the key feature of Nord
ctricity tricty Trans- Pool. Major contractual relation among Nordic countries
Commis- Commis- mission is given in figure 2.
sion)genera sion) System
ting com- generat- Authori-
panies and ing com- ty) Gen-
SMHEA(Sn panies eration
owy moun- plus corpora-
tain hydro SMHEA tion trad-
electricity ing as
authority) optima
energy.
Trans- Transmis- Trans- Trans- Transmission
mission sion commission mission company.
wires pany:Trans compa- compa-
grid ny:Power ny:ETSA
Net Vic- Trans-
toria mission
corp.
Bulk NSW Re- Victoria Partici- State
Fig2. Nordic Market Major Contractual Relationship
Market gion of region of pating in poll,separate
NEM1 NEM1 the Vic- market com-
2.5.1 Nord Power (Pool Group)
torian pany
Region of
NEM1. (i) Nord Pool Spot :
Distribu- 6 distribu- 5 distrib- ETSA 7 distribu-
tion tors with utors power tion wires
It consist of Nord Pool Spot AS and its wholly owned
wires ring-fenced with corpora- business
subsidiary Nord Pool Finland Oy, operates the physical
retailers ring- tion
day-ahead market Elspot in whole Nordic region and the
fenced
physical intra-day market Elbas in Finland, Sweden and
retailers
Zealand (Eastern Denmark). Elspot and Elbas are Nord
Retail 6 host re- 5 host ETSA 3 host retail-
Pool Spot auction based markets for trade in power con-
Supply tailers and retailers power ers and un-
tracts for physical delivery. On Elspot, hourly power con-
unlimited and un- corp.(hos limited
tracts are traded daily for physical delivery in the next
indepen- limited t retail- supply li-
day's 24-hour period. On Elbas, continuous adjustment
dent supply supply ers);unli censes
trading in hourly contracts can be performed until one
licenses licenses mited
hour before the delivery hour. Its function is to be the
supply
aftermarket to the Elspot market at the Nord Pool.
licenses
(ii) Nord Pool ASA - Financial Market :
Table1.Electricty Restructuring in Australia
Nord Pool Financial market is a regulated market
2.5 Nordic Power Market place which trade in standardized derivative instruments
like forward and future contracts going out several years,
and has now started trade in options. Outside this mar-
In Norway, the electricity reforms were initiated in 1991.
ket, there is quite a large and liquid market for over the
In 1993, Nordic power exchange was established as an
counter forward and option contracts. The objective of
IJSER 2011
89
ISSN 2229-5518
financial market is to provide an efficient market, with are given the freedom and responsibility of controlling
excellent liquidity and a high level of security to offer a (scheduling) their resources, and have to optimize the
number of financial power contracts that can be used utilization of their physical and contractual as-
profitably by a variety of customer groups. This market is sets.Transmission system operations are organized on a
wholly owned by Nord Pool Group. national basis for Nordic countries.The Five TSOs in the
Nordic area are owner of respective main national grid
(iii) Nord Pool Clearing ASA : The National Transmission System Operators (TSOs) are
responsible for reliability and balance settlements.
It is a licensed and regulated clearing-house. It is cen- The Elspot market is formed as a day-ahead physical-
tral counter party for all derivative contracts traded delivery power market and the deadline for submitting
through exchange and OTC. It guarantees settlement for bids for the following days delivery hours is fixed as 12
trade and anonymity for participants. It is wholly owned am (noon). There are three types of bids available in Els-
subsidiary of Nord Pool Group. pot; the hourly bid or single bid, block bid and flexible
hourly bid.Participants can submit bids to Nord Pool Spot
(iv) Nord Pool Consulting AS : electronically either through EDIEL communication or
through the internet application ElwWeb.Nord Pool of-
It is a consulting firm specializing in development of fers futures contracts for one to nine days ahead and for
power market worldwide. It is also a wholly owned sub- one to six weeks ahead in time. These futures contracts
sidiary of Nord Pool Group. are settled daily. All these futures and forward contracts
use the daily average system price as reference. There are
(v) Nordel: also contracts to hedge zonal price differences, either one
quarter or one year ahead.Prices for real-time are deter-
Nordel is an association for electricity cooperation mined by the marginal bid like in the day-ahead spot
between forum for market participants, nordic system market.Real time market is also known as Regulated
operators and TSOs of nordic countries. The primary ob- Power Market in Norway.Nord Pool PX has a market
jectives of organization are to create and maintain the share of 43% of the physical Nordic demand; the remain-
necessary conditions for an effective nordic electricity ing 57% is traded bi-laterally.Nord Pool also operates a
market. trading platform for financial derivatives as well as clear-
ing house for bi-lateral contracts.
Congestion Management is done by market splitting
i.e. resolving congestion in day ahead market and counter
trade i.e. resolving congestion in real time.Point of Con-
2.5.2 Nord pool Features nection tariff structure is followed to promote space.UI
pricing mechanism is followed for deviation from sche-
Nord pool is first multinational commodity exchange dule.Dr. Per Christer,Senior Vice President ,Nord Pool
for Electric sector in the world .It provide open market to Consultancy has given the Nord Pool Market Model
all Nordic Countries with common framework.-Nordic (shown in fig3.).
and Europian markets are example of decentralized day
ahead spot market.There are no general cross border tariff
among Nordic Countries.Trading of electricity generated
by hydropower dominates the cross-border exchanges
between the Nordic countries. The balance of electricity
trade between the four countries depends on rainfall con-
ditions because of great variation in fuel type capacity of
Nordic countries. If hydropower potential is good, Swe-
den and Norway record trade surpluses, if hydro re-
source is poor Denmark and Finland will benefit from the
electricity trading.-There are only one Market Operator
(MO)-Nord Pool and five System Operator (SO) which
are Svenska Kraftnt in Sweden, Fingrid in Finland, Stat-
nett in Norway, Eltra in western Denmark, and Elkraft
System in eastern Denmark[9].There are separate regula- Figure 3.Nord Pool Model as given by Dr. Per Christer
tory agencies in the four countries. The MO is in principle
only responsible for facilitating the trade of electricity as
the commodity, but within the physical constraints set by 2.6 Pennsylvania New Jersey Maryland (PJM)
the SO. The operation of the physical system is the sole Market
responsibility of the SO. Further, the market participants The Pennsylvania New Jersey Maryland in-
IJSER 2011
90
ISSN 2229-5518
terconnection (PJM) has been a pool between the Figure 4. PJM Model
three founding utilities that enables co-ordination of
trade since 1927. PJM is responsible for the manage- The Comprehensive National Energy Strategy an-
ment of a competitive wholesale electricity market nounced on April 1998 and stressed that it relies as
across the control areas of its members and for safe much as possible on free markets and competition.
and reliable operation of the unified transmission sys- An Independent System Operator (ISO) and Power
tem. All generators defined as a capacity resource in Exchange (PX) have been established in 1998 based
PJM system are obliged to submit an offer into the on market structure after the CPUCs decision in 1995
day-ahead PJM market. Market participants are al- which became watershed for the road towards com-
lowed to self schedule. Transmission system security petitive market. The outlining of the proposed Cali-
and reliability considerations are taken into account fornia Model filed with FERC on April 29, 1996 with-
for the total market clearing operation. A marginal three investor owned utilities (IOUs) in California
pricing principle is used for market clearing. Each which were Pacific Gas & Electric or PG&E, Southern
generator at its specific node is paid market clearing California Edison and San Diego Gas Electric. There
price . All loads at their specific nodes are charged as are three significant characteristic in California Mod-
per the market-clearing price . PJM Interconnection is el-
a non-profit company ,a limited liability, governed by a) To simplify the transmission pricing scheme in-
a board of managers. There is a specific unit Market cluding nodal and congestion charge assessing, Zonal
Monitoring Unit (MMU) within PJM to oversee the Approach is applied.
functioning of the market. States have public utility b) A Scheduling coordinator (SC) or PX have been
commissions (PUCs) and the Federal Energy Regula- introduced to manage multiple separate energy for-
tory Commission (FERC)(shown in fig4). PUCs regu- ward markets (each with a supply and demand port-
late generation and distributions intra-state utility folio). An adjustment approach is adopted to perform
business. The FERC regulates interstate energy trans- inter-zonal congestion management.
actions including wholesale power transactions on c) An adjustment bid approach is adopted to per-
transmission lines[10]. form inter-zonal congestion management.
An Independent System Operator (ISO) and a
2.7 California State Power Exchange (PX) have been established in 1998
based on market structure and rules governed by
Public, Political pressure and higher electricity cost FERC.Multiple separate energy forward markets,
have resulted in ending the regulated monopolies of
each with a supply and demand portfolio managed
vertically integrated utilities. Deregulation in US pro-
ceeded with the Public Utility Regulating Policies Act by a Scheduling Coordinator (SC) or PX have been
approval in 1978 and the Energy Policy Act (EPAct) introduced .The total separation of the wholesale
in 1992.Federal Energy Regulatory Commis- power exchange and the market participant was done
sion(FERC) approved non-discriminatory open from ISO.
access to transmission services in 1995.Utilities and Power Exchange will be independent entitiy for
Regulators, including American Electric Power (AEP) managing bid of energy for each half-hour on a day
, the California Public Utilities Commission ahead basis for ISO dispatch decision. The ISO will
(CPUC),the New England Electric System (NEES) and
the Pennsylvania/New Jersey/Maryland(PJM) pool
control the power dispatch and the transmission sys-
have formulated several proposal for change[11]. tem [12]. It will have no financial interest in the Pow-
er Exchange or in any generation, load, and
transmission or in distribution facilities. The
ISO will coordinate the information exchange
in an open market and willwork as per North
American Reliability Council (NERC) and
Western System Coordinating Council (WSSC)
reliability standards. The ISO will coordinate
day-ahead scheduling and balancing for all us-
ers of the transmission grid and also will pro-
cure ancillary services.
IJSER 2011
91
ISSN 2229-5518
Scheduling Coordinators (SCs) aggregate partici- and a spot market for electricity.The EMCO operates
pants in the energy trade and are free to use protocols the market througha bidding system and is the clear-
that may differ from pool rules. SCs run a forward ing house for market transactions.TransPower , the
market in which parties can bid to buy and sell ener- operator and developer of the national grid ,performs
gy and submit the preferred schedule to the ISO and the various services like provision of reliable national
work with the latter to adjust schedules when neces- grid ,efficient scheduling and dispatch generation to
sary(figure 5). satisfy market demand,purchasing of ancillary ser-
vices and providing information to the grid users in
an open ,non-discriminatory manner in the whole-
sale electricity market.
2.9 Canada market

California PX
The restructuring in the Albertas electric industry
was started to retain benefits of the existing low cost
generators for customers in 1996.In the new structure
,power pool was defined for all energy- trade in the
province.Generation sector made fully competitive
with competitive bidding except the case of old re-
tired power plants.IPP was brought to meet load
growth. Grid Company of Alberta,(GridCo) adminis-
Figure 5.California structure with inclusion of Scheduling tered a province wide transmission grid.The trans-
Coordinators.
mission grid was owned by the four utilities that own
transmission facilities in the province and contract all
2.8 New Zealand Electricity market
individual owners to supply transmission servic-
es.The Electric Transmission Council as advisory
The New Zealand electricity system consists of
group formed to represent the interests of consumers
Nort and South Island as two alternative current (AC
and transmission users[14].
system connected by 1200 MW underwater HVDC
cable.All capacity on South Island is hydroelectric
3.0 England & Wales market
and it exported power to North Isl-
and.Approximately 75% of the North Island demand
UK government took a historic step by privatizing
is met from hydroelectric source with the remaining
the publicly owned electric power industry in Eng-
25% split between geothermal sources and fossil fuel
land and Wales (E&W) in 1988.Generation, Transmis-
(coal, gas and oil) sources. Prior to February 1,1996
sion and Distribution of electricity were divided into
,the generating industry was dominated by state-
three large companies. All existing fossil fuel plants
owned Electricity corporation of New Zeal-
were taken over by National Power and powerGen .
and(ECNZ) which owned and operated 95% of total
The National Grid Company (NGC) provides
capacity.A wholesale market as Contact Energy ltd.
transmission services from generators to the Regional
for electricity was formed as seperate state owned
Electricity Supply Companies (RECs) and coordinates
enterprise from ECNZ On February 1,1996. There are
transmission and dispatch of electricity generators
currently 38 electricity distribution compa-
[15].NGC runs both the physical and financial side of
nies,providing equal access distribution services and
the E& W electricity market. It serves as both the In-
electricity supply to customers and one electricity
dependent system operator (ISO) and the Power Ex-
retailer providing electricity supply only[13].The
change and determine both half-hourly market clear-
wholesale electricity market in New Zealand com-
ing prices and runs the physical national grid, mak-
menced operation under the name Electricity Market
ing generator dispatch decisions in real time to man-
Company(EMCO) on October 1,1996.
age congestion on the grid and provide ancillary ser-
The New Zealand Electricity Market(NZEM) intro-
vices for reliable supply to all the consumers [16].
duces competition within the wholesale electricity
NGC uses GOAL (generation ordering and loading)
sector through creation of a national electricity pool
IJSER 2011
92
ISSN 2229-5518
program to determine the merit order of dispatching demand of SEB through respective RLDCs at regional
generation alongwith reserve capacity[17]. level. But it is mandatory cost based (non bid) Power
Pool.ABT mechanism facilitates Balancing market in
3.1 Indian Market an inherent way but it has got some limitations al-
so.Current trading occur between ISGS and states
Indian Power sector is in a transition phase from a STU/SEBS,
regulated sector to a competitive market (taken as
author is from India). A competitive market provides
State
the participants with benefits of psrice determination
Genco
by market forces ,easy access to market, transparent CGS
working however it also brings with it many changes
that need to be taken care of by the market partici-
pants at various stages of development.
The power sector in India has seen significant de-
velopments post the enactment of the Electricity Act
2003[18]. The policy and regulatory efforts have also
been synchronized to ensure rapid development of
the power markets in the country.
In this direction, Electricity Act 2003 has come into
force from June 2003 in India. It introduces the con-
cept of trading bulk electricity. The Act has enabled
consumers and the distribution companies to have
choice in the selection of electricity supplies . Similar- Figure 6.Indian Market Structure after Act-2003
ly, the generator also has choice to select among the
between states, through international import
distribution companies (shown in figure 6.). The Act
/export (Bhutan, Nepal) and also by state embedded
specifies the provisions for non-discriminatory use of
transmission lines or distribution system or asso- generators/IPPs / Loads and others.The RLDCS or-
ganize the day ahead scheduling of the ISGS[20]
ciated facilities with such lines or system by any li-
censee or consumer or a person engaged in genera- .Short term bilateral contracts are taking place
through traders but they are lacking formal market
tion.
and real time information. Often Sellers call for sepa-
At regional level, there are five regional load dis-
rate tenders for surplus available with them and trad-
patch centers( NR,WR,ER,SR,NER) which are operat-
ers compete with each others on prices to get the
ed by Power Grid..At state level, there are 28 states
supply. This situation has resulted in prices of traded
which are responsible for their generation, transmis-
sion and distribution. States purchase power from power moving only in one direction (higher). The
root cause is one-sided competition. On the other
Independent State Generation Supply (ISGS).Trade
hand, buyers are not getting adequate response
between states is facilitated by trading firms like
PTC,NVVL and others[19].Distribution licenses and against tenders called by them. A platform for wide
sellers and buyers is not available.
Government do not need trading license and trans-
mission licenses and load dispatch centers can not
4 CONCLUSION
trade power.About 2.5 % of total power generated in
country is being traded presently.There are 17 li-
In the various countries, most of the electric power
censed electricity traders for inter state trading till
industry has been going through a process of transi-
now.Most generation capacity (56%
tion and restructuring by moving away from vertical-
State,36%CGS,11%Private) tied up with long term
ly integrated monopolies and towards more competi-
contracts. Only surplus can be traded.The present
tive market models since nineties. This has been
inter-regional capacity is 11500MW which is planned
achieved through as creating competition at each lev-
to go upto 37000 MW upto 2012.India have Pool type
el in the power industry and having a clear separa-
centralized mechanism for dispatching central gene-
tion between its generation, transmission and distri-
rating plants. On day ahead basis to meet forecast
bution activities as well. Different countries are im-
IJSER 2011
93
ISSN 2229-5518
plementing industry restructuring in a variety of

ways, depending on the characteristics of each mar-
ket area which include: diversity of generation by
fuel types, demand/supply balances, the extent of
transmission capacity to facilitate energy imports to
meet market demand and etc.. In designing and
planning the market structure and rules for competi-
tion in their jurisdiction, governments, regulators and
other industry participants are influenced by local
market characteristics and the practices.
References
[1] M,Illic, F. Galiana and L. Fink, Power System Restructur-
ing:Engineering and Economics,Kluwer Academy Publisher,1998.
[2] http://class.ece.iastate.edu.
[3] A.K.Izaguirre, Private participation in the electricity sector-
recent trends,World Bank report-Private Sector, December
1998,pp.5-12.
[4] H. Rudnick , Chile: Pioneer in deregulation of the electric
power sector,IEEEE Power Engineering Review ,June 1994,pp.
28-30.
[5] M.I.Dussan ,Restructuring the electric power sector in Colom-
bia,IEEE Power Engineering Review,June 1994,pp21-22.
[6] C.M. Bastos, Electric energy sector in Argentina,IEEE Power
Engineering Review June 1994 ,pp.13-14.
[7] H. Outhred, A review of electricity industry restructuring in
Australia,Electric Power Systems Research,vol.44,1998,pp.15-
25.
[8] R.D.Christie and I. Wangesteen,The energy market in Norway
and sweden:Introduction,IEEE Power Engineering Re-
view,February 1998,pp.44-45.
[9] http://www.nordpool.com.
[10] PJM Interconnection LLC,PJM Open Access Transmission
Tariff: Schedule-2, Fourth Revised, Vol1, Issued February
2001.
[11] Z.Alaywan and J.Alen,California electric restructuring ;a
broad description of the development of the California
ISO,IEEE Trans on Power Systems ,Vol.13,No.4,November
1998,pp.1445-1451.
[12] http://www.caiso.com.
[13] T.Alvey,D.Goodwin,X. Ma, D. Streiffert and D Sun,A security
constrained bid-clearing system for the New Zealand wholesale elec-
tricity market,IEEE Trans on Power Sys-
tems,Vol.13,No.2,May1998,pp.340-346.
[14] J.R. Frey,Restructuring the electric power industry in Alberta
,IEEE Power Engineering Review ,February 1996,pp.8.
[15] National Grid Electricity Transmission (NGET) plc, The connection
and use of system code (CUSC), Issued Feb.2006.
[16] http://www.nationalgrid.com/uk.
[17] http://www.iitk.ac.in.ime/anoops.
[18] http://www.cerc.org.
[19] http://www.ee.iitb.ac.in/wiki/faculty/sak.
[20] http://www.jbic.go.jp/en/research/report.
IJSER 2011
94

ISSN 2229-5518
Significant Role of Search Engine in

Higher Education
Rahul J. Jadhav, Dr. Om Prakash Gupta, Usharani T. Pawar
Abstract- Information explosion has given a rise to quest for more and more knowledge and its applicability to varied fields. Higher education is
not exception to this. The facilities, of varied institutions are perpetually involved in teaching, learning, evaluation and research activities. They are
using the modern technology to a great extent. Not only the faculty and academicians but students of todays world also use the latest technology
for knowing more and more. Therefore the use of internet has taken rapid stride. For collecting data and information varied programs are
developed and the use of search engine prove to be the most significant tool for gathering information and knowledge. Search engine is one of the
most widely used method for navigating of cyberspace. The objective of the research paper is to study the significant role of search engine to
make the higher education innovative and easily accessible to the students, faculty and researchers.
Keywords: Search engine, Higher education, navigating of cyberspace.
1. INTRODUCTION documents and creates an index based on the words

Search Engine: A Capsule Description contained in each document. Each search engine uses a
proprietary algorithm to create its indices such that,
A web search engine is designed to search for ideally, only meaningful results are returned for each
information on the World Wide Web and FTP servers. query.
The search results are generally presented in a list of
results and are often called hits. The information may
consist of web pages, images, information and other 1.2. History of Search Engine
types of files. Some search engines also mine data
available in databases or open directories. Unlike Web "How could the world beat a path to your door when the
directories, which are maintained by human editors, path was uncharted, uncatalogued, and could be discovered
search engines operate algorithmically or are a mixture only serendipitously?" Paul Gilster, Digital Literacy.
of algorithmic and human input.
History of Search Engine can be said as started in A.D.
1.1. What is Search Engine? 1990. The very first tool used for searching on the
A program that searches documents for specified Internet was Archie (The name stands for "archives"
keywords and returns a list of the documents where without the "v", not the kid from the comics). It was
the keywords were found. Although search engine is created in 1990 by Alan Emtage, a student at McGill
really a general class of programs, the term is often University in Montreal. The Archie Database was
used to specifically describe systems like Google, Alta made up of the file directories from hundreds of
Vista and Excite that enable users to search for systems. When you searched this Archie Database on
documents on the World Wide Web and USENET the basis of a files name, Archie could tell you which
newsgroups. directory paths on which systems hold a copy of the
Typically, a search engine works by sending file you want. Archie did not index the contents of
out a spider to fetch as many documents as possible. these sites. This Archie Software, periodically reached
Another program, called an indexer, then reads these out to all known openly available ftp sites, list their
files, and build a searchable index. The commands to

search Archie were UNIX commands, and it took some
Bharati Vidyapeeth Deemed University,Pune Yashwantrao Mohite Institute
knowledge of UNIX to use it to its full capability.
of Management, Karad. INDIA E-mail [email protected]
Bharati Vidyapeeth Deemed University,Pune Yashwantrao Mohite Institute
of Management, Karad. INDIA
Two other programs, "Veronica" and
Shivaji University, Kolhapur Department of computer science S.G.M College "Jughead," searched the files stored in Gopher index
, Karad. INDIA E-mail [email protected] systems. Veronica (Very Easy Rodent-Oriented Net-
IJSER 2011
95

ISSN 2229-5518
wide Index to Computerized Archives) provided a This problem might be considered to be a mild form of
keyword search of most Gopher menu titles in the linkrot, and Google's handling of it increases usability
entire Gopher listings. Jughead (Jonzy's Universal by satisfying user expectations that the search terms
Gopher Hierarchy Excavation And Display) was a tool will be on the returned webpage. This satisfies the
for obtaining menu information from various Gopher principle of least astonishment since the user normally
servers. expects the search terms to be on the returned pages.
Increased search relevance makes these cached pages
In 1993, MIT student Matthew Gray created very useful, even beyond the fact that they may contain
what is considered the first robot, called World Wide data that may no longer be available elsewhere.
Web Wanderer. It was initially used for counting Web
servers to measure the size of the Web. The Wanderer When a user enters a query into a search
ran monthly from 1993 to 1995. Later, it was used to engine (typically by using key words), the engine
obtain URLs, forming the first database of Web sites examines its index and provides a listing of best-
called Wandex. matching web pages according to its criteria, usually
with a short summary containing the document's title
In 1993, Martijn Koster created ALIWEB and sometimes parts of the text. The index is built from
(Archie-Like Indexing of the Web). ALIWEB allowed the information stored with the data and the method
users to submit their own pages to be indexed. by which the information is indexed. Unfortunately,
According to Koster, "ALIWEB was a search engine there are currently no known public search engines
based on automated meta-data collection, for the Web." that allow documents to be searched by date. Most
1.3. How Search Engine Works? search engines support the use of the Boolean operators
AND, OR and NOT to further specify the search query.
A search engine operates, in the following order Boolean operators are for literal searches that allow the
user to refine and extend the terms of the search. The
Web crawling
engine looks for the words or phrases exactly as
Indexing entered. Some search engines provide an advanced
feature called proximity search which allows users to
Searching define the distance between keywords. There is also
concept-based searching where the research involves
Web search engines work by storing information about using statistical analysis on pages containing the words
many web pages, which they retrieve from the html or phrases you search for. As well, natural language
itself. These pages are retrieved by a Web crawler queries allow the user to type a question in the same
(sometimes also known as a spider) an automated form one would ask it to a human. A site like this
Web browser which follows every link on the site. would be ask.com.
Exclusions can be made by the use of robots.txt. The
contents of each page are then analyzed to determine 2. Importance of Search Engine
how it should be indexed (for example, words are The usefulness of a search engine depends on the
extracted from the titles, headings, or special fields relevance of the result set it gives back. While there
called Meta tags). Data about web pages are stored in may be millions of web pages that include a particular
an index database for use in later queries. A query can word or phrase, some pages may be more relevant,
be a single word. The purpose of an index is to allow popular, or authoritative than others. Most search
information to be found as quickly as possible. Some engines employ methods to rank the results to provide
search engines, such as Google, store all or part of the the "best" results first. How a search engine decides
source page (referred to as a cache) as well as which pages are the best matches, and what order the
information about the web pages, whereas others, such results should be shown in, varies widely from one
as AltaVista, store every word of every page they find. engine to another.
This cached page always holds the actual search text
since it is the one that was actually indexed, so it can be In cyberspace, there's no place to "turn." I have
very useful when the content of the current page has only my computer screen in front of me. Somehow, I
been updated and the search terms are no longer in it. need to find a place to purchase the book I want.
IJSER 2011
96

ISSN 2229-5518
There's no street on my screen so I can't drive around 3. Occasionally, they find a site by hearing about it
on the Web (I could "surf," but that's hit and miss; even from a friend or reading in an article.
then I still need to know where to start). Sometimes it's Thus its obvious that the most popular way to find a
obvious: type in the name of the bookstore, add a site is search engine.
.COM and it's a pretty good bet you're going to end up
where you want to go. But what if it's a specialty Table 1.1 Top Ten Search Engines
bookstore and doesn't have a Web site with an obvious
URL? Search Engine No. of Percentage
Respondents
One solution to this problem is the search Google.com 112 57.5
engine. In fact, it's probably one of the most widely Yahoo.com 39 19.5
used methods for navigating in cyberspace.
Considering the amount of information that's available Bing.com 13 6.5
from a good search engine, it's similar to having the Ask.com 11 5.5
Yellow Pages, a guide book and a road map all-in-one. AOL.com 9 4.5
AltaVista.com 7 3.5
Search engines can provide much more Alltheweb.com 4 2
information than just the URL of a Web site. Typing in Lycos.com 3 1.5
"books" into the Google search engine returns about Excite.com 1 0.5
9,270,000 results. If we refine the search to "books, HotBot.com 1 0.5
Internet", we end up with about 6,070,000 results. If we
know the book's author, let's say E.Balguruswamy
books , search engines now returns About 80,500
Top Ten Search Engines
results within 0.18 seconds (of course, these results will
No. Respondents(%)
change from day to day).For many people, using search

70
engines has become routine. Are the Search Engines are 60
important? Undoubtedly, positively, absolutely....YES! Here's 50
how important they are. In the recent Georgia Tech Internet 40
Series1
30
User Survey, respondents were asked how they find pages 20
on the Internet. A full 82% said they used the major search 10
engines. 0
Google.com
Yahoo.com
Bing.com
Ask.com
AOL.com
HotBot.com
AltaVista.com
Alltheweb.com
Lycos.com
Excite.com
It is the search engines that finally bring your website
to the notice of the prospective customers. When a
topic is typed for search, nearly instantly, the search
engine will sift through the millions of pages it has
indexed about and present you with ones that match Search Engine
your topic. The searched matches are also ranked, so
that the most relevant ones come first.
It is the Keywords that play an important role than any As you can see from the statistics, Google absolutely
expensive online or offline advertising of your website. dominates the search engine market. Its closest
It is found by surveys that a when customers want to competitor is Yahoo.com but they seem to be endlessly
find a website for information or to buy a product or buying old search technologies that do not provide any
service, they find their site in one of the following innovative techniques. This bodes well for Googles
ways: continued dominion.
1.The first option is they find their site through a
4. Role of Search Engine in Higher
search engine.
2. Secondly they find their site by clicking on a link
Education
from another website or page that relates to the topic in To conduct an effective search, the researcher must
which they are interested. understand the structure of the various search engines.
Search engines do not always provide the right
IJSER 2011
97

ISSN 2229-5518
information, but rather often subject the user to a almost 4 hours a day online (see table 1.3) and the
deluge of disjointed irrelevant data. majority of that time is spent at work (see table1.4)
All search engines support single-word Table 3 Daily Time spent online
queries. The user simply types in a keyword and
presses the search button. Most engines also support Daily Time spent online Percent
multiple-word queries. However, the engines differ as
to whether and to what extent they support Boolean < 1 hour 10.4
operators (such as "and" and "or") and the level of 1-2 hours 24.8
detail supported in the query. More specific queries 2-3 hours 33.1
will enhance the relevance of the user's results. 3+ hours 31.1
4.1. Variations on the Search Engine Table 4 Work VS Personal Internet Use by the
Respondents
A search engine is not the same as a "subject directory."
A subject directory does not visit the web, at least not Work VS Personal Internet Use Percent
by using the programmed, automated tools of a search By The Respondents
engine. Websites must be submitted to a staff of trained 0% personal/ 100%work 7.8
individuals entrusted with the task of reviewing,
50% personal/ 50%work 68.6
classifying, and ranking the sites. Content has been
screened for quality, and the sites have been
categorized and organized so as to provide the user
with the most logical access. Their advantage is that
Respondents were asked the first place they
they typically yield a smaller, but more focused, set of
would go online to learn more about the product or
results.
service they were considering. Search was the clear
Table 2 Search result. winner over manufacturers sites and information
portals, with 66.3 % of respondents. (see Table 1.5).
Bharati
Notes on Indian
Keyword Vidyapeeth Table 5 first place to find out educational
JAVA Railway
information
Google 56,000,000 272,000
2,580,000
Where would be the first place Percent
Yahoo 59,600,000 293,000 you would go online to find out
17,200,000
educational information
Bing 8,41,00,000 Search Engine 66.3
16,00,000 208,000
Independent Web site 21.6
AltaVista 92,800,000
17,300,000 64,200 Educational Portal 8.3
AOL 3,820,000 Other 4.8
2,210,000 95,100
With the majority of respondents indicating that search

Above table 1.2 shows, no search engine covers the
plays a major role in their education, we next asked
entire web. There are technical obstacles such as the
which engine they would use to launch their search.
inability to index frames, image maps, or dynamically
We fully expected Google to be the winner, but we
created websites
were surprised by how much they dominated their
4.2. Importance of Search in higher education
competition (Table 1.6). An amazing 90.9% chose
There can be no downplaying the importance of search
Google as their engine of choice.
in higher education. Search continues to be the number
one method for finding relevant information online. Table 6 Search engine chosen by the Respondents
Respondents indicated that they spend an average of
Search engine chosen by the Percent
IJSER 2011
98

ISSN 2229-5518
Respondents
Google 90.9
Yahoo 4.7
MSN 4.3
AltaVista 0.3
Lycos 0.2
5. Conclusion
Table 7 Conclusion
Daily Time spent online
< 1 hour 1-2 hours 2-3 hours 3+ hours

10.4 24.8 33.1 31.1
Work VS Personal Internet Use by the Respondents
0% 50% 75% 100%

personal personal/ personal/ personal/ 6. References
/100%work 50%work 25%work 0%work
7.8 68.6 21.2 2.4 [1] Craig Lerner. The importance of search engine. Uber Articles
First place to find out educational information March 2, 2010
[2] Broder, A. Taxonomy of web search, SIGIR Forum, vol. 36,pp. 3-
10, 2002.
Search Independent Educationa Other [3] Web search engine - Wikipedia, the free encyclopedia
Engine Web site l Portal
[4] Are Search Engines Important? Searchengineposition.com
66.3 21.6 8.3 4.8
[5] How Do Web Search Engines Work webopedia 02-17-2006
Search engine chosen by the Respondents
[6] LevelTen_Colin The role of Search Engine Optimization in
Internet Marketing http://www.articlesbase.com/
Google Yahoo MSN AltaVista [7] The Role of Search in B2B Buying Decisions White paper -
90.9 4.7 4.3 0.3 Enquiro Search Solutions
[8] Lee Underwood A Brief History of Search Engines
[9] Brooks, N. 2004. The Atlas Rank Report II: How searchengine
This study revealed information that will be key to rank impacts conversions. Atlas Institute, 2004.
formulating effective search in higher education. This [10] Finkelstein, L., Gabrilovich, E., Matias, Y., Rivlin, E., Solan,Z.,
study also reveled that majority of the respondents Wolfman G., and Ruppin, E. Placing search in context: the concept
search for information on general search engines like revisited. Proceedings of the WWW Conference, 2001.
[11] Greenspan, R. Searching for balance. vol. 2004: ClickZ stats.
Google, yahoo. Google is overwhelming the search
[12] R. Villa, M. Chalmers, A framework for implicitly
engine. Percentage of respondents who search for trackingdata, Proceedings of the Second DELOS Network of
information relevant to their objective is very low. ExcellenceWorkshop on Personalisation and Recommender Systems
inDigital Libraries, Dublin City University, Ireland, June 2001.
[13] iProspect Inc. Search engine user attitudes, 2005.
[14] Jansen, B., and Spink, A. The effect on click-through
ofcombining sponsored and non-sponsored search engine
results in a single listing, Proceedings of the 2007 Workshop
on Sponsored Search Auctions, WWW Conference, 2007.
IJSER 2011
99
ISSN 2229-5518
Feature Selection for Cancer Classification: A

Signal-to-noise Ratio Approach
Debahuti Mishra, Barnali Sahu
Abstract Cancers are generally caused by abnormalities in the genetic material of the transformed cells. Cancer has a reputation as a
deadly disease hence cancer research is intense scientific effort to understand disease. Classification is a machine learning technique used
to predict group membership for data instances. There are several classification techniques such as decision tree induction, Bayesian
classifier, k-nearest neighbor (k-NN), case-based reasoning, support vector machine (SVM), genetic algorithm etc. Feature selection for
classification of cancer data is to discover gene expression profiles of diseased and healthy tissues and use the knowledge to predict the
health state of new sample. It is usually impractical to go through all the details of the features before picking up the right features. This paper
provides a model for feature selection using signal-to-noise ratio (SNR) ranking. Basically we have proposed two approaches of feature
selection. In first approach, the genes of microarray data is clustered by k-means clustering and then SNR ranking is implemented to get top
ranked features from each cluster and given to two classifiers for validation such as SVM and k-NN. In the second approach the features
(genes) of microarray data set is ranked by implementing only SNR ranking and top scored feature are given to the classifier and validated.
We have tested Leukemia data set for the proposed approach and 10fold cross validation method to validate the classifiers. The 10fold
validation result of two approaches is compared with hold out validation result and again with results of leave one out cross validation
(LOOCV) of different approaches in the literature. From the experimental evaluation we got 99.3% accuracy in first approach for both k-NN
and SVM classifiers with five numbers of genes and with 10fold cross validation method. The accuracy result is compared with the accuracy
of different methods available in the literature for leukemia data set with LOOCV, where only multiple-filter-multiple wrapper approach gives
100% accuracy in LOOCV with leukemia data set.
Index TermsClassification, Feature selection, Cancer data, Microarray, Signal-to-noise ratio
1 INTRODUCTION
A LL organisms except viruses consist of cells. East has

one cell, where as human have trillions of cells.
acquired mutations and epigenetic changes that influence
gene expression. A major focus in cancer research is
Document cell consists of nucleus and inside nucleus identifying genetic markers. Clinical diagnosis of cancer
there is DNA, which encodes the programs for making based on gene expression data has two main targets: first
future organisms. Genes make proteins in two steps. First to achieve the correct diagnostic for a cancer patient with
DNA transcribed to mRNA and mRNA is translated in to a greatest confidence. Second, to identify the gene
proteins [1]. Gene expression is the activation of genes responsible for a particular type of cancer, this helps in
that results in a protein. Proteins are the blue prints for the diagnosis and prognosis of cancer. These objectives
the characteristics of the living organisms. imply to develop best classification models which ensure
A microarray is a sequence of dots of DNA, protein, or a true classification of a cancer sample with a low risk of
tissue arranged in an array for easy simultaneous misclassification. Many high level data analysis
analysis. The most famous is the DNA microarray, which techniques such as clustering and classification
plays an integral role in gene expression profiling. The algorithms work better with smaller number of genes.
substrate material is glass, plastic or a silicon chip. This approach usually covers one or more components of
Important applications of microarrays include the microarray data analysis that include dimensionality
identification of genetic individuality of tissues or reduction through a gene subset selection, the
organisms, the diagnosis of genetic and infectious disease construction of new predictive features and model
[2][3]. inference [2].
The goal of this paper is to make an intensive study on
Debahuti Mishra, ITER, Siksha O Anusansdhan University, Odisha, India. the techniques available for finding the patterns among
Email: [email protected]
Barnali Sahu, ITER, Siksha O Anusansdhan University, Odisha, India. the genes or feature selection using SNR ranking and to
Email: [email protected] analyze the result of our two approaches for feature
selection which gives significant meaning to classify the
genes which are responsible for cancer disease.
Cancers are caused by abnormalities in the genetic This paper is arranged in the following way:
materials of the transformed cells. It mostly results from introduction to cancer classification data is given in
IJSER 2011
100
ISSN 2229-5518
section 1, section 2 deals with preliminary concept of pairs of data items. The major advantage of k-NN is its
microarray, classification techniques, SNR ranking, k- simplicity.
means clustering. Section 3 deals with related work on
feature selection of cancer data using SNR approach, Support Vector Machine (SVM): Support vector machines
section 4 deals with the proposed model, section 5 (SVM) is a supervised learning techniques which analyze
contains experimental evaluation, section 6 explains the data and recognize patterns, used for statistical methods
validation and comparison of our work and section 7 and regression analysis[7]. SVM training algorithm builds
concludes the paper. a model that predicts whether a new sample falls into one
category or the other. SVM model is a representation of
the samples as points in space, mapped so that the
2 PRELIMINARIES
samples of the separate categories are divided by a clear
2.1 Microarray gap that is as wide as possible. New samples are then
All cells in an organism carry the same genetic mapped into that same space and predicted to belong to a
information and only a subset of the genes is active category based on which side of the gap they fall on.
(expressed). Analyzing the gene with respect to whether Support vector machine constructs a hyper plane or a set
and to what degree they are expressed can help of hyper planes in a high or infinite dimensional space,
characterize and understand their functions. It can further which can be used for classification, regression or other
be analyzed how the activation level of genes changes tasks.
under different conditions such as for specific diseases
[3][4]. 2.3 k-means clustering Algorithm
Microarray data are generally high dimensional data
having large number of genes in comparison to the
number of samples or conditions. There are many Input: k = Number of clusters
efficient methods for the analysis of microarray data such P= A data set containing n features (n
as clustering, classification and feature selection. number of genes)
Feature selection is the preprocessing task for both
clustering and classification. Different types of 1. Select number of cluster k.
experiment can be done by microarray technology. 2. Randomly choose k features from the data set as
Microarray technology measures the expression level of the initial cluster center.
genes. That can be used in the diagnosis, through the 3. Repeat until the termination criteria fulfilled
classification of different types of cancerous genes leading 3.1 Assign each feature to one of the clusters
to a cancer type[5].Basically, genes of microarray data are according to the similarity measure
treated as features, a set of features(genes) give rise to a 3.2 Update the cluster means.
pattern. If we could get the correct pattern from the data 4. until no change in the value of clusters mean
set it is easier to classify an unknown sample based on
that pattern. In this approach we have used Euclidean distance as
distance measure.
2.2 Classification Technique Revisited
Our study is mainly based on feature selection and 2.4 Signals-to-Noise Ratio
pattern classification for gene expression data related to
cancer diagnosis. There are several classification The signal to noise ratio (SNR) test identifies the
techniques such as SVM, k-NN, neural network, nave expression patterns with a maximal difference in mean
bayesian, decision tree, random forest, top scoring pair. expression between two groups and minimal variation of
expression within each group [8]. In this method genes
k-NN: k-NN is the simplest ML technique for classifying are first ranked according to their expression levels using
objects based on closest training examples in the feature SNR test Statistic. The SNR is defined as follows:
space[6]. It is instance based learning. It gathers all Signal to noise ratio= (1 +/2 )( 1+ 2) (1)
training data and classifiers often via a majority voting, a
new data point with respect to the class of its k-nearest Where 1 and 2 denote the mean expression values
neighbor in the given data set. k-NN obtain the neighbors for the sample class 1 and class 2 respectively. 1 and 2
in the given data set. k-NN obtain the neighbors for each
are the standard deviations for the samples in each class.
data by using Euclidian or Mahalanobis distance between
IJSER 2011
101
ISSN 2229-5518
3 RELATED WORK
Wai-Ho et al.[9] presents an attribute clustering method
Apply SNR ranking
,which is able to group genes based on their Select top ranked
to clusters
interdependence to mine meaningful patterns from genes
microarray data. Gene selection methods used are
Attribute clustering, t-value, k-means, Biclustering, Collect top scored
MRMR, RBF and Classifiers used are C5.0, Neural genes from clusters
Networks, Nearest Neighbor, Nave Bayes. Data sets used
for the experiment are Colon cancer and Leukemia. Supoj
Train the classifiers
Hengpraprohm et al.[10] proposed a method which
with filtered genes
yields higher accuracy than using the SNR ranking alone
and higher than using all of the genes in classification.
Selection of informative features using k-means and Validation
SNR ranking. DLBCL, Ovarian, Colon, Prostrate, Breast
cancer, CNS, Leukemia, Lung Cancer are the data sets
used for the experiments. Hualong Yu et al. [11]
Accuracy Accuracy
demonstrated that a modified discrete PSO is a useful tool
for selecting marker genes and mining high dimensional
Fig.1. Model for the comparison of accuracies of SVM and kNN in
data.
two approaches with 10 fold cross validation method
SNR ranking is used to select top ranked informative
genes. Then PSO is applied to select few marker
genes.SVM is used for evaluation of prediction. Colon
cancer data set is used for the experiment. Yukee Leung et
5 EXPERIMENTAL EVALUATION
al.[12] make use of multiple filter and multiple wrappers
to improve the accuracy of the classifiers. We have used leukemia data set of cancer microarray
data from Biological data analysis web site [21]. The Data
Some of MFMW selected genes have been conformed
set contains 7,129 genes and 72 samples (47 ALLs, 25
to be biomarkers. Multiple filters are SNR, Pearson
AMLs). For our approach we have taken 50 genes and 72
correlation, t-statistics. Multiple wrappers are SVM, WV, samples (47 class1, 25 class2) of original data set. The
3NN and data sets used are LEU [13], COL62 [14], experiment is done in MATLAB version 7.6.0.324
BRER49 [15], LYM77 [16], PROS102 [17], LUNG182 [18]. (R2008a), windows XP, PC of Intel Pentium dual CPU.
Shamsul Huda et.al. [19] proposed a hybrid wrapper and We have implemented two different approaches of
filter feature selection algorithm by introducing filters feature selection used for classification model to discover
feature ranking score in wrapper stage to get a more differentially expressed genes.
compact feature set. They have hybridized mutual
information based maximum relevance filter ranking 5.1 First Approach for Feature Selection
method with artificial neural network based wrapper
approach to get the accuracy. Chenn-jung Huang et.al [20] Step 1: First, the features of data are clustered by
Have under gone a comprehensive study on the applying k-means clustering algorithm. As by
capability of probabilistic neural network associated with applying clustering technique we can group
SNR scoring method for cancer classification. The similar type of features in same cluster so that
experimental results show that the combination of PNN best feature from each cluster can be selected. In
with the SNR method can achieve better results for our approach we have tested the model with 5,
Leukemia data set. 10 and 20 clusters.
Step 2: The features in each cluster is ranked by

4 PROPOSED MODEL
applying signal-to-noise ratio scoring technique,
so that differentially expressed genes can be
Microarray easily extracted from each cluster.
Data Step3: After that best scored feature in each cluster is
selected. We can assure that applying SNR and
selecting the best scored feature from each cluster
Cluster genes by k- Apply SNR ranking the resultant feature gene subset have no
means to feature set
IJSER 2011
102
ISSN 2229-5518
redundancy. different classifiers independently.

Step4: The data with the selected biomarkers are tested Step5: 10fold cross validation accuracy is listed in table
by different classifiers. The classifiers used are 2.
3NN, SVM
Step5: For validation we have used 10fold cross TABLE 2
validation approach. The performance of ACCURACY OF SVM AND K-NN IN SECOND METHOD AND SVM
different classifiers with respect to the number of
clusters is given in table 1
TABLE 1 10fold CV
No of
Method Data set accuracy
ACCURACY OF SVM AND K-NN IN FIRST METHOD WITH genes
(%)
DIFFERENT CLUSTERS
5 97.5
SNR+SVM Leukemia 10 96.1
From the above table 1 we can see that both SVM and 20 91.4
kNN classifiers are giving same accuracy with 5 numbers
5 95.4
of genes in 10fold cross validation method i.e 99.3%. The
SNR+ k-NN Leukemia 10 90.0
comparison of accuracy of two classifiers are given bellow
20 98.1
in fig.2
From table 2 we can observe that kNN classifier with 20

genes give better result is 98.1% accuracy than SVM
classifier in 10fold cross validation approach. The
10fold CV accuracy chart of second approach is given bellow in fig.
Method No of
Data set accuracy 2
clusters
(%)
5 99.3
Kmeans+SNR+ SVM Leukemia 10 94.1
20 96.1
5 99.3
Kmeans+SNR+ kNN Leukemia 10 89.3
20 94.9
Fig.3. Accuracy of SVM and kNN in second approach with 10fild

cross validation
In the paper [22] we have implemented the given two

approaches with leukemia data set and validated with k-
NN, SVM, PNN and FNN classifiers. We have used hold
Fig.2. Accuracy of SVM and kNN in first approach with 10fold cross
out validation method to find the accuracy of different
validation
classifiers with the above two feature selection approach.
In this paper we have considered only two classifiers such
as SVM and k-NN the accuracies of these two classifiers
5.2 Second Approach for Feature Selection with hold out validation method for two approaches are
given in table 3 and 4 respectively.
Step1: SNR scoring technique is applied to the rows of
the data set. TABLE 3
Step2: Basing on the SNR score the features are ranked. HOLD OUT VALIDATION ACCURACY OF SVM AND K-NN IN
Step3: 5, 10, 20 top scored features are chosen
FIRST METHOD
randomly.
Step4: The new data set with these features are fed to
IJSER 2011
103
ISSN 2229-5518
Hold out
No of validation
Method Data set
clusters accuracy
(%)
5 100
Kmeans+ SNR+
Leukemia 10 96
SVM
20 96
5 96
Kmeans+ SNR+
Leukemia 10 83
kNN
20 87
Fig.5. Accuracy of k-NN in first approach with hold out
TABLE 4
HOLD OUT VALIDATION ACCURACY OF SVM AND K-NN IN and 10fold cross validation
SECOND METHOD
Hold out
Method Data set No of validation
genes (%)
SNR+SVM Leukemia 5 96
10 96
20 96
SNR+kNN Leukemia 5 96
10 96
20 96
Fig.6. Accuracy of k-NN in second approach with hold out

and 10fold cross validation
Fig.4. Accuracy of SVM in first approach with hold out and

10fold cross validation Fig.7. Accuracy of SVM in second approach with hold out
and 10fold cross validation
From fig. 4 we can analyze that SVM classifier gives

better result in hold out validation method than 10fold 6 COMPARISON AND VALIDATION
cross validation method. But from figure 5 we can
conclude that kNN classifier gives better result in first Comparing fig.2 and fig.3 we can analyze that the 10fold
approach with 10fold cross validation method. cross validation accuracy of k-NN and SVM is same i.e
99.3% which is better than the second approach. Where in
the second approach SVM gives 97.5% with 5 genes and
IJSER 2011
104
ISSN 2229-5518
k-NN gives 98.1% accuracy with 20 genes .Hence in

second approach k-NN gives better result with large
number of features
In case of hold out validation method only SVM
gives 100% accuracy with and k-NN gives 96% accuracy
with 5 numbers of genes.
From the above comparisons we can conclude that
SVM gives better results with less number of features in
first approach. Fig.8. Accuracy of different approaches for Leukemia data
Now if we are going to compare the results of two set with LOOCV
validation method for our two approaches than from fig.
4 and fig.5 we can see that SVM gives better result in hold
out validation than 10fold cross validation method for
7 CONCLUSION
first approach. But k-NN gives better result for 10fold
cross validation than hold out validation. From the above comparative analysis we can conclude
From fig.6 and fig.7 we can analyze that in second that our first approach for feature selection is better in
approach both k-NN and SVM gives better accuracy in comparison to our second approach as due to clustering
10fold validation method. technique similar features will be grouped in to the same
From the literature we have collected the accuracies of clusters. After applying SNR ranking and selecting top
different methods or approaches for LOOCV method for scored features from each cluster may give a true pattern
Leukemia data set, given in table 5 and the accuracy chart which helps to enhance the classification accuracy. But in
of different approaches are given in fig.8. case of second approach after applying SNR ranking we
Again comparing our approach and methods present can randomly choose the top scored features where we
in literature we can see that MFMW gives 100% accuracy can get redundant feature or noisy features with similar
with 4 numbers of genes and our first approach with SNR score and does not provide any relevant information
Hold out validation with SVM gives 100% accuracy but about the data. Therefore the performance of the learning
first approach with 10fold cross validation approach in algorithms decreases. Again in comparison to hold out
SVM and k-NN classifier gives 99.3% accuracy with 5 validation and 10fold validation approach k-NN and SVM
numbers of genes. perform well in first approach for 10fold cross validation
and only SVM perform well for hold out validation
TABLE 5 method.
LOOCV ACCURACY OF DIFFERENT APPROACHES FOR
LEUKEMIA DATASET
REFERENCES
Method Accuracy(%)LOOCV
MFMW[16] 100
[1] Gregory Piatetsky-Shapiro, Pablo Tamayo, Microarray
MLP+SNR[11] 76.5 Data Mining: Facing the Challenges, SIGKDD Explorations,
Volume5, Issue 2, pp. 1-5, June 2003
SVM(linear)+SNR[11] 58.8 [2] Minca Mramor Gregor Leban, Janez Demar and Bla
Zupan, 2007, "Visualization-based cancer microarray
kNN(Pearson)+SNR[11] 97.1 data classification analysis", Bioinformatics, Vol. 23,
No.16, pp.2147-2154, 2007.
GPC+ clus[11] 90.3 [3] Wolfgang Huber, Anja Von Hey debreck, Martin
Vingron,Analysis of microarray gene expression data, Hand
book of statistics genetics, 2nd edition, Wiley.2003
[4] Hong-Hai Do,Toralf Kirsten,Erhard Ralm,Comparative
Evaluation of Microarray-based Gene expression Database,GI-
Proceedings, pp 26-34.
[5] Ana C.lorena, Ivan G.costa, Marcilio c. p. de Souto,On the
complexity of gene expression classification data sets, Eighth
International Conference on Hybrid intelligent System,pp
825-830.2008
[6] V.N. Vapnik, Statistical Learning Theory, Wiley-
Interscience Publications, 1998
IJSER 2011
105
ISSN 2229-5518
[7] Vapnik VN.The nature of statistical Theory.Springer- Research, vol. 62, no. 17, pp. 4963-4967, 2002.
Verlag;1995 [19] Shamsul Hunda, John Yearwood, Andrew Strainieri,
[8] Miroslava Cuperlovic-Cuf, Nabil Belacel, Rodney. j. Hybrid wrapper-filter approach for input feature
Ouellette, Determination of Tumour marker genes from gene selection using Maximum Revalance and Artificial Neural
expression data, DDT, Vol-10, Number 6 pp429-437, 2005 Network Input Gain Measurement Approximation,
[9] Wai-Ho Au,Keith C.C.Chan,Andrew K.C. Wong, Yang Fourth International conference on Network and system
Wang. Attribute clustering for Grouping, IEEE/ACM security, pp442-449, 2010.
Transactions on computational biology and [20] Chenn-Jung Huang ,Wei-Chen Liao, A Comparative
Bioinformatics, Vol 2.,No 2, pp83-101,2005 Study of Feature Selection Methods for Probabilistic
[10] Supoj Hengpraprohm, Prabhas Chongstitvatana, Selecting Neural Networks in Cancer Classification, Proceedings of
Informative Genes from Microarray Data for Cancer the 15th IEEE International Conference on Tools with Artificial
Classification with Genetic Programming Classifier using K- Intelligence (ICTAI03),Vol 3, pp1082-3409, 2003.
Means Clustering and SNR Ranking, Frontiers in the [21] http://sdmc.lit.org.sg/GEDatasets/
Convergence of Bioscience and Information Technologies , [22] Debahuti Mishra, Barnali Sahu,A signal to noise
pp211-216, 2007. classification model for identification of differentially expressed
[11] Hualong Yu,Guochang Gu,Haibo Liu,Jing Shen, genes from gene expression data,3rd International conference
Changming Zhu,. A Novel Discrete Particle Swarm on electronics computer technology,2011(Accepted)
Optimization Algorithm for Microarray Data-based Tumor
Marker Gene Selection, International Conference on Computer
science and software Engineering, pp. 1057-1060, 2008
[12] Yukyee Leung, Yeungsam Hung, A Multi-Filter-Multi-
Wrapper Approach to Gene Selection and Microarray Date
Classification, IEEE/ACM Transactions on Computational
Biology and Bioinformatics, Vo1. 7, No .1, pp.108-117, 2010.
[13] T.R. Golub, D.K. Slonim, P. Tamayo, C. Huard, M.
Gaasenbeek, J.P. Mesirov, H. Coller, M.L. Loh, J.R.
Downing, M.A. Caligiuri, C.D. Bloomfield, and E.S.
Lander, Molecular Classification of Cancer: Class Discovery
and Class Prediction by Gene ExpressionMonitoring, Science,
vol. 286, no. 5439, pp. 531-537, 1999.
[14] U. Alon, N. Barkai, D.A. Notterman, K. Gish, S. Ybarra, D.
Mack,and A.J. Levine, Broad Patterns of Gene Expression
Revealed by Clustering Analysis of Tumor and Normal Colon
Tissues Probed by Oligonucleotide Arrays, Proc. Natl
Academy of Sciences USA, vol. 96, no. 12, pp. 6745-6750,
1999
[15] M. West, C. Blanchette, H. Dressman, E. Huang, S. Ishida,
R. Spang, H. Zuzan, J.A. Olson Jr., J.R. Marks, and J.R.
Nevins, Predictingthe Clinical Status of Human Breast
Cancer by Using Gene Expression Profiles, Proc. Natl
Academy of Sciences USA, vol. 98, no. 20, pp. 11462-11467,
2001.
[16] M.A. Shipp, K.N. Ross, P. Tamayo, A.P. Weng, J.L. Kutok,
R.C.T. Aguiar, M. Gaasenbeek, M. Angelo, M. Reich, G.S.
Pinkus, T.S. Ray, M.A. Koval, K.W. Last, A. Norton, T.A.
Lister, J. Mesirov, D.S. Neuberg, E.S. Lander, J.C. Aster,
and T.R. Golub, Diffuse Large B-Cell Lymphoma Outcome
Prediction by Gene-Expression Profiling and Supervised
Machine Learning, Nature Medicine, vol. 8, pp. 68-74,
2002.
[17] D. Singh, P. Febbo, K. Ross, D. Jackson, J. Manola, C. Ladd,
P. Tamayo, A. Renshaw, A. DAmico, and J. Richie, Gene
Expression Correlates of Clinical Prostate Cancer Behavior,
Cancer Cell, vol. 1, no. 2, pp. 203-209, 2002.
[18] G.J. Gordon, R.V. Jensen, L.L. Hsiao, S.R. Gullans, J.E.
Blumenstock, S. Ramaswamy, W.G. Richards, D.J.
Sugarbaker, and R. Bueno, Translation of Microarray Data
into Clinically Relevant Cancer Diagnostic Tests Using Gene
Expression Ratios in Lung Cancer and Mesothelioma, Cancer
IJSER 2011
106
ISSN 2229-5518
Towards Analysis of Vulnerabilities and

Measuring Security Level of Copmputer
Network Life Cycle
Irshad Ahmad Mir, Mehraj-U-Din Dar, S.M.K Quadri
Abstract To measure the overall security of a network, a crucial issue is to correctly compose the measure of individual
components. Incorrect compositions may lead to misleading results. The detection of vulnerabilities and security level estimation
are the main and important tasks of protecting computer networks. To obtain correct compositions of individual measures, we
need to first understand the interplay between network components. For example, how vulnerabilities can be combined by
attackers in advancing an intrusion. This paper takes into consideration the models and architecture of intelligent components
intended for analyzing computer network vulnerabilities and costing the security level of the computer networks
Index Terms Security Metrics, Vulnerability Assessment, Attack Model, Security level Assessment.
.
1 INTRODUCTION
C RUCIAL to todays economy and national security,

computer networks plays a central role in most en-
terprises and critical infrastructures. In protecting
Therefore now vulnerability detection and estimation of
security level of computer networks are actual tasks of
information assurance. A special class of systems exists
these networks against malicious intrusions, a standard for the solution of these tasks vulnerability assessment or
way for measuring network security is utmost important. security analysis systems (SAS) [17, 20]. The contempo-
Despite existing efforts in standardizing security metrics rary SAS destine to fulfill checking the system defended
[16, 27], a widely accepted network security metrics is against the specified system configuration and security
largely unavailable. At the research frontier, a qualitative policy for non-compliance and identifying technical vul-
and imprecise view toward the evaluation of network nerabilities in order to correct them and mitigate any risk
security is still dominant. According to CERT statistic [1] posed by these vulnerabilities. The main objective of SAS
the quantity of attacks on computer networks, their com- components is to identify and correct the system man-
plexity and extent of damage, caused by malefactors at- agement process and security policy failures that pro-
tacks in the Internet, grows each year. The reason is a low duced the vulnerabilities detected. The other important
security level of majority of systems connected to the In- functions are security level estimation, supporting effec-
ternet. The most common failures exist in operating sys- tive interface for control of scanning process, creating
tem (OS) and applications software configuration, soft- reports and automatic updating vulnerability signatures.
ware maintenance, user management and administration, The SAS components should scan system, update the sys-
including improperly configured OS and applications, tem configuration according to the specified security poli-
incorrect password policy and improper access control cy and system configuration and also send inquiries to
settings, existence of vulnerable or easily exploited ser modify the security policy if it is necessary. It is also ne-
vices and malicious software (Trojans, worms, etc.). cessary to carry out vulnerability assessment and security
analysis during the whole life cycle of computer net-
works, including initial stage of analysis and design.
Irshad Ahmad Mir is currently pursuing Ph.D degree program in comput- In this paper we focus on creating the models, architec-
er science department in University of Kashmir , India..He did his bache- ture and prototypes of intelligent components of vulnera-
lors degree in computer application from Amar Singh College Srinagar In-
dia and Master degree in computer application from Kashmir university bility detection and security level estimation by expand-
India. ing the functional capabilities of existing SAS based on
E-mail: [email protected]. simulation and penetration testing. We mainly focus on to
design stage. The rest of the paper is structured as fol-
Dr. Mehraj-U-Din Dar is currentlythe director of Information Technolo-
gy & Support Systems at University of Kashmir , India. lows. Section 2 outlines the approach suggested and re-
lated work. Section 3 describes the architecture of security
Dr. SMK Quadri is Head, PG department of computer sciences, Kashmir analysis system developed and its implementation issues.
University , India. He did his Mtech in computer application from Indian
school of Mines and Ph.D in computer sciences from Kashmir University ,
Section 4 gives an outline of generalized attack model
India. used for vulnerability assessment and security level esti-
mation. Section 5 describes the model of analyzed com-
IJSER 2011
107
ISSN 2229-5518
puter network. Section 6 gives an overview of Experimen- ate on the stage of exploitation. Examples are NetRecon,
tal approach. Section 7 Draws the conclusion. bv-Control for Internet Security (HackerShield), Retina,
Internet Scanner, CyberCop Scanner, Nessus Security
2. RELATED WORK Scanner,
etc. The basic lacks of existing SAS are as follows: (1) use
At the design stage, SAS should operate with the model of the scanner does not allow to answer to the main ques-
of analyzed computer Network generated from prelimi- tion concerning policy-based systems - Whether what is
nary or detailed design specifications. The main ap- revealed during scanning correspond to security policy?;
proaches to vulnerability assessment and security analy- (2) the quality of obtained result essentially depends on
sis can be based on analytic calculation and imitation (si- the size and adequacy of vulnerability bases;
mulation) experiments. Analytical approaches use as a (3) Implementation of active vulnerability analysis on the
rule different risk analysis methods [2, 11, 22, 24, 30, etc.]. computer system functioning in a regular mode can lead
Imitational approaches are based on modeling and simu- to failures in running applications. Therefore not all sys-
lation of network specifications, fault (attack) trees, graph tems can be tested by active vulnerability analysis.
models, etc. [9, 10, 11, 13, 17, 21, 32, 26, 27, 28, 31, etc.].
There are a lot of papers which consider different tech-
niques of attack modeling and simulation: Colored Petri 3. ARCHITECTURE OF SECURITY ANALYSIS SYSTEM
Nets [15], state transition analysis technique [12, 14], si- Architecture of security analysis system (SAS) given
mulating intrusions in sequential and parallelized forms (fig.1) contains following components: (1) user interface
[5], cause-effect model [6], conceptual models of comput- (2) module of malefactors model realization (3) mod-
er penetration [29], descriptive models of the network and ule of script set (attack scenarios) generation (4) module
the attackers [33], structured tree-based description [7, of scenario execution (5) data and knowledge repository
19], modeling survivability of networked systems [18], (6) module of data and knowledge repository updating
object-oriented discrete event simulation [3], requires/ (7) module of security level assessment (8) report genera-
provides model for computer attacks [32], situation calcu- tion module (9)network interface.
lus and goal-directed procedure invocation [8], using and
building attack graphs for vulnerability analysis [13, 31],
etc.
As one can see from our review of relevant works, the
field of imitational approaches for vulnerability assess-
ment and security level evaluation has been delivering
significant research results. [25] Quantifies vulnerability
by mapping known attack scenarios into trees. [16] Sug-
gests a game-theoretic method for analyzing the security
of computer networks. The authors view the interactions
between an attacker and the administrator as a two-
player stochastic game and construct a model for the
game. The approach offered in [34] is intended for per-
forming penetration testing of formal models of net-
worked systems for estimating security metrics. The ap-
proach consists of constructing formal state/transition
models of the networked system. The authors build ran-
domly constructed paths through the state-space of the
model and estimate global security related metrics as a
function of the observed paths. [31] Analyzes risks to spe-
cific network assets and examines the possible conse-
quences of a successful attack. As input, the analysis sys-
tem requires a database of common attacks, specific net-
work configuration and topology information, and an
attacker profile. Using graph methods they identify the
attack paths with the highest probability of success. [10]
Suggests global metrics which can be used to analyze and
proactively manage the effects of complex network faults
Fig.1. Generalized Architecture SAS
and attacks, and recover accordingly.
There are a lot of different SAS components which oper-

IJSER 2011
108
ISSN 2229-5518
The knowledge base about analyzed system in- of interaction with a computer network transferring,
cludes data about the architecture and particular parame- capturing and the preliminary analysis of network traffic.
ters of computer network (for example, a type and a ver- The preliminary analysis includes: (1) parsing of packets
sion of OS, a list of opened ports, etc) which are needed according to connections and delivery of information
for scripts generation and attack execution. This data about packets (including data on exposed flags, payload,
usually can be received by malefactor using reconnais- etc.) and connections; (2) acquisition of data about attack
sance actions and methods of social engineering. results and system reactions, and also values of some sta-
The knowledge base of operation (functionality) tistics reflecting actions of SAS at the level of network
rules contains Meta and low-level rules of IF-THEN packets and connections.
type determining SAS operation on different levels of The module of security level assessment is based
detail. Meta-level rules define attack scenarios on higher on developed taxonomy of security metrics. It is a main
levels. Low level rules specify attack actions based on module which calculates security metrics based on results
external vulnerability database. IF-part of each rule con- of attack actions. The module of database and knowledge
tains (meta-) action goal and (or) condition parts. The repository update downloads the open vulnerability da-
goal is chosen in accordance with a scenario type, an at- tabases (for example, OSVDB - open source vulnerability
tack intention and a higher level goal (specified in a meta- database) and translates them into KB of operation (func-
rule of higher level). The condition is compared with the tionality) rules of low level
data from database about analyzed system. THEN-part
contains the name of attack action which can be applied
4. GENERALIZED ATTACK MODEL
and (or) the link on exploit. An example of one of rules is
IF GOAL = Denial of service AND OS_TYPE = Win- The model is defined as hierarchical structure that con-
dows_XP AND OS_VERSION =4 THEN ping_of_death sists of several levels as shown in fig.2. Three higher le-
(PoD). Each rule is marked with an identifier which al- vels of the attack model correspond to an attacks script-
lows us to determine the achieved malefactors goal. For set, a script and script stages. The scriptset level defines a
example, the rule mentioned above defines a denial of set of general malefactors intentions. The second script
service (DoS) attack ping_of_death. level defines only one malefactors intention. Third the set
The DB of attack tools (exploits) contains ex- of script stages can contain the following elements: re-
ploits and parameters of their execution. A choice of a connaissance, implantation (initial access to a host), gain-
parameter is determined by the data in KB about ana- ing privileges, threat realization, covering tracks and
lyzed system. For example, the program of ftp brute force backdoors creation. Lower levels serve for malefactor sub
password cracking needs to know the ftp server port goals refinement. The lowest level describes the malefac-
which can be determined by port scanning. tors low level actions directly executing different ex-
The module of scriptset (attack scenarios) gen- ploits.
eration selects the data about analyzed system from the
data and knowledge repository, generates attack scriptset
based on using operation (functionality) rules, monitors
scriptset execution and scriptset updating at runtime,
updates data about analyzed system.
The module of scenario execution selects an at-
tack action and exploits, prognoses a possible feedback
from analyzed computer network, launches the exploit
and recognizes a response of analyzed computer network.
In case of interaction with a computer network a real
network traffic is generated. In case of operation with the
model of analyzed system two levels of attack simulation
are provided: (1) at the first level each low-level action is
represented by its label describing attack type and (or)
used exploit, and also attack parameters; (2) at the second
(lower) level each low-level action is specified by corres-
ponding packets of the network, transport and applied
level of the Internet protocols stack.
Network interface provides: (1) in case of opera-
tion with the model of analyzed system transferring
identifiers and parameters of attacks (or network packets
under more detailed modeling and simulation), and also
receiving attacks results and system reactions; (2) in case Fig.2. Generalized attack Model
IJSER 2011
109
ISSN 2229-5518
5. ANALYZED NETWORK MODEL consists of notions of attack realization action, as well as the
notions of type and categories of assets. There are four levels
The analyzed computer network model plays a role in
of security metrics sub-taxonomy based on attack realization
evaluating attack results and defining the reaction of the
actions (fig. 4) (1) an integrated level; (2) a script level; (3) a
system. It contains following basic components as shown
level of the script stages; (4) a level of the threat realization.
in fig.4 network interface, module of malefactors action
Each higher level contains all metrics of lower levels (arrow
recognition; module of attack result evaluation; module in fig.4 shows the direction of metrics calculation). Examples
of system response generation; database of analyzed sys- of security metrics for this taxonomy are as follows: number
tem; database of attack signatures. of total and successful attack scenarios; number of total and
Network interface provides: (1) receiving iden- successful stages of attack scenarios; number of total and
tifiers and parameters of attacks; (2) transferring attack successful malefactor attacks on the certain level of taxono-
results and system reactions. my hierarchy; number of attacks blocked by existing security
Module of malefactor actions recognition is ne- facilities; number of discovered and used vulnerabilities;
cessary at realization of detailed attack modeling and si- number of successful scenario implementation steps; num-
mulation, i.e. when malefactor actions are represented as ber of different path of successful scenario implementation,
network packets. Functioning of this module is based on etc.
a signature method the data received from the network
interface are compared to signatures of attacks from data-
base of attack signatures. Outputs of the module are iden-
tifiers and parameters of attacks.
The knowledge base about analyzed system is
created from the specification of analyzed system and
structurally coincides with KB about analyzed system
described in section 2. The difference of these knowledge
bases consist in the stored data: KB of the model of ana-
lyzed system contains the results of translating the speci-
fications of analyzed system; KB related to the genera-
lized architecture of SAS is initially empty and is filled
during the execution of attack scripts.
Fig. 4. Security Metrics sub taxonomy Based on Attack

Realization action
6. EXPERIMENTAL APPROACH
We had conducted test based on simulation. The experi-
ment is carried out using OPNET MODELER 14.0
which allows us simulate and to form a virtual computer
networks. In our test the network consists of three sub-
nets :
Internet area including hosts Int_host and
Fig.3. Model of Analyzed Computer Network ISP_DNS with ip adress 192. 17. 300.*
A logical sub network including one server with
IP address 192.168.0.*.
5. SECURITY LEVEL ESTIMATION MODEL
Local area network with IP addresses 100.0.0.*.
The security level evaluation is described by multi-level The basic components (Fig.5) of network are : (1) Internet
hierarchy of security metrics . The taxonomy of security me- host with integrated system of software. ;(2) Firewall 1
trics is based on the attack model developed. This taxonomy between logical subnetwork and the Internet ; (3) Mail
IJSER 2011
110
ISSN 2229-5518
server; (4) File server ;(5) Firewall 2 between logical sub- According to the malefactors model realization SAS
network and LAN ; (6) A Local DNS server, services the creates one script consisting of the following two stages: (1)
clients from LAN; (7) An authentication, , authorization reconnaissance and (2) threat realization (denial of service).
and accounting server; (8) Workstations 1 and 2 . We now We calculate the confidentiality and criticality levels of
consider the SAS prototype for our experiment. We de- successfully attacked assets. At reconnaissance stage, the
termine the level of the file-server against attacks denial malefactor has received the information which total level of
of service taking into consideration that the malefactors confidentiality is 10 and total level of criticality is 4. For the
experience is low . To do this we need to count up the information which the malefactor tried to receive the ap-
assets and its criticality levels (CRL) and confidentiality propriate levels are (20, 20). After normalization, the losses
levels(COL) . Table.1. shows the CRL and COL of neces- of confidentiality and criticality are (0.6, 0.3). At thread rea-
sary assets. lization stage, the file-server has been successfully attacked
(0 points of confidentiality and 10 points of criticality have
been lost), therefore the appropriate losses are (0, 1). At
script level the losses of confidentiality and criticality are as
follows: ((0.6+0)/2, (0.3+1)/2) = (0.3, 0.65). The total securi-
ty metric can be calculated as difference 1 and average val-
ue of the given coefficients: 1-0.475=0.525. Let us select by
expert evaluation the following security level scale: (1)
green if security level value in an interval [1, 0.8); (2)
yellow [0.8, 0.6); (3) red [0.6, 0]. Then the value
0.525 acts as red level. As guideline on increase of securi-
ty level, the report about vulnerability elimination is gen-
erated. Procedure of security level evaluation is repeated
after eliminating detected vulnerabilities.
7. CONCLUSION
The paper offered the approach to vulnerability analysis and
security level assessment of computer networks, intended
for implementation at various stages of computer network
This work is based on basic components of intelligent SAS,
the model of computer attacks and the model of security
level assessment based on developed taxonomy of security
metrics. The SAS prototype was implemented and the expe-
riments were held based on simulation of attacks.
Fig. 5.Configuration of computer Network for Experiment
REFERENCES
[1] CERT/CC Statistics 1988-2005.
http://www.cert.org/stats/cert_stats.html
[2] Chapman, C., Ward S.: Project Risk Management: processes, techniques
and insights. Chichester, John Wiley (2003)
[3] Chi, S.-D., Park, J.S., Jung K.-C., Lee J.-S.: Network Security Modeling
and Cyber Attack Simulation Methodology. LNCS, Vol.2119 (2001)
[4]
[5] Chung, M, Mukherjee, B., Olsson, R.A., Puketza, N.: Simulating Con-
current Intrusions for Testing Intrusion Detection Systems. Proc. of the
18th NISSC (1995)
[6] Cohen, F.: Simulating Cyber Attacks, Defenses, and Consequences.
IEEE Symposium on Security and Privacy, Berkeley, CA (1999)
[7] Dawkins, J., Campbell, C., Hale, J.: Modeling network attacks: Extend-
ing the attack tree paradigm. Workshop on Statistical and Machine
Learning Techniques in Computer Intrusion Detection, Johns Hopkins
Table 1. COR and COL of Assets University (2002)
[8] Goldman R.P.: A Stochastic Model for Intrusions. LNCS, V.2516 (2002)
IJSER 2011
111
ISSN 2229-5518
[9] Gorodetski, V., Kotenko, I.: Attacks against Computer Network: For- Attacks. Proc. Of the New Security Paradigms Workshop (2000).
mal Grammar-based Framework and Simulation Tool. RAID 2000. [33] Yuill, J., Wu, F., Settle, J., Gong, F.: Intrusion-detection for incident-
LNCS, V.2516 (2002) response, using a military battlefield-intelligence process. Computer
[10] Hariri, S., Qu, G., Dharmagadda, T., Ramkishore, M., Raghavendra C. Networks, No.34 (2000).
S.: Impact Analysis of Faults and Attacks in Large-Scale Networks.
IEEE Security & Privacy, September/ October (2003)
[11] Henning, R.: Workshop on Information Security System Scoring and
Ranking. Williamsburg, VA: Applied Computer Security Associates and
The MITRE Corporation (2001)
[12] Iglun, K., Kemmerer, R.A., Porras, P.A.: State Transition Analysis: A
Rule-Based Intrusion Detection System. IEEE Transactions on Software
Engineering, 21(3) (1995).
[13] Jha, S., Linger, R., Longstaff, T., Wing, J.: Survivability Analysis of Net-
work Specifications. bIntern. Conference on Dependable Systems and
Networks, IEEE CS Press (2000).
[14] Kemmerer, R.A., Vigna, G.: NetSTAT: A network-based intrusion
detection approach. 14th Annual Computer Security Applications Confe-
rence, Scottsdale, Arizona (1998)
[15] Kumar, S., Spafford, E.H.: An Application of Pattern Matching in Intru-
sion Detection. Technical Report CSDTR 94 013. Purdue University
(1994)
[16] Lye, K., Wing J.: Game Strategies in Network Security. International
Journal of Information Security, February (2005)
[17] National Institute of Standards and Technology. Technology assess-
ment: Methods for measuring the level of Computer security. NIST
Special Publication 500-133, 1985.
[18] Moitra, S.D., Konda, S.L.: A Simulation Model for Managing Surviva-
bility of Networked Information Systems, Technical Report CMU/SEI-
2000-TR-020, December (2000)
[19] Moore, A.P., Ellison, R.J., Linger, R.C.: Attack Modeling for Information
Security and Survivability. Technical Note CMU/SEI-2001-TN-001.
March (2001)
[20] M. Swanson, N. Bartol, J. Sabato, J. Hash, and L. Graffo. Security metrics
guide for information technology systems. NIST Special Publication
800-55, 2003.
[21] Nicol, D.M., Sanders, W.H., Trivedi, K.S.: Model-Based Evaluation:
From Dependability to Security. IEEE Transactions on Dependable and
Secure Computing. Vol.1, N.1 ( 2004)
[22] Peltier, T.R.: Information security risk analysis. Auerbach 2001.
[23] Peltier, T.R., Peltier, J., Blackley, J.A.: Managing a Network Vulnerability
Assessment. Auerbach Publications (2003)
[24] RiskWatch users manual. http://www.riskwatch.com
[25] Schneier, B.: Attack Trees. Dr. Dobbs Journal, vol. 12 (1999)
[26] Sheyner, O., Haines, J., Jha, S., Lippmann, R., Wing, J.M.: Automated
generation and analysis of attack graphs. Proc. of the IEEE Symposium on
Security and Privacy (2002)
[27] Singh, S., Lyons, J., Nicol, and D.M.: Fast Model-based Penetration
Testing. Proceedings of the 2004 Winter Simulation Conference (2004)
[28] Steffan, J., Schumacher, M.: Collaborative Attack Modeling. 17th ACM
Symposium on Applied Computing (SAC 2002), Madrid, Spain (2002)
[29] Stewart, A.J.: Distributed Metastasis: A Computer Network Penetration
Methodology. Phrack Magazine, 9 (55) (1999)
[30] Storms A.: Using vulnerability assessment tools to develop an OC-
TAVE Risk Profile. SANS Institute.http://www.sans.org
[31] Swiler, L., Phillips, C., Ellis, D., Chakerian, S.: Computer-attack graph
generation tool. DISCEX, 2001
[32] Templeton, S.J., Levitt, K.: A Requires/Provides Model for Computer
IJSER 2011
112

ISSN 2229-5518
Text Independent Speaker Identification In a

Distant Talking Multi-microphone
Environment Using Generalized Gaussian
Mixture Model
P. Soundarya Mala, Dr. V. Sailaja, Shuaib Akram
Abstract -- In speaker Identification System, the goal is to determine which one of the groups of an unknown voice which best matches with
one of the input voices. The field of speaker identification has recently seen significant advancement, but improvements have tended on near
field speech, ignoring the more realistic setting of far field instrumented speakers. In this paper, we use far field speech recorded with multi
microphones for speaker identification. For this we develop the model for each speakers speech. In developing the model, it is customary to
consider that the voice of the individual speaker is characterized with Generalized Gaussian model. The model parameters are estimated using
EM algorithm. Speaker identification is carried by maximizing the likelihood function of the individual speakers. The efficiency of the proposed
model is studied through accuracy measure with experimentation of 25 speakers database. This model performs much better than the existing
earlier algorithms in Speaker Identification.
Keywords-- Generalized Gaussian model, EM Algorithm, and Mel Frequency Cepstral Coefficients.
INTRODUCTION
when close talking microphones are used. In adverse
distant-talking environments, however the performance
S peaker recognition is the process of recognizing
who is speaking on the basis of information
is significantly degraded due to a variety of factors such
as the distance between the speaker and microphone, the
extracted from the speech signal. It has been number of
location of the microphone or the noise source, the
applications such as verification of control access
direction of the speaker and the quality of the
permission to corporate database search and voice mail,
microphone. To deal with these problems micro phone
government lawful intercepts or forensics applications,
arrays based speaker recognizers have been successfully
government corrections, financial services, telecom & call
applied to improve the identification accuracy through
centers, health care, transportation, security, distance
speech enhancement [3][4][6].
learning, entertainment & consumer etc [2].
In speaker identification since there is no identity claim,
the system identifies the most likely speaker of the test
The growing need for automation in complex work
speech signal. Speaker identification can be further
environments and increased need for voice operated
classified into closed-set identification and open-set
services in many commercial areas have motivated for
identification. Speaker identification can be further
recent efforts in reducing laboratory speech processing
classified into closed-set identification and open-set
algorithms to practice. While many existing systems for
identification. The task of identifying a speaker who is
speaker identification have demonstrated good
known a priori to be a member of the set of N enrolled
performance and achieve high classification accuracy
speakers is known as closed-set speaker Identification.
The limitation of this system is that the test speech signal
*Ms P.Soundarya mala Completed M.tech in Digital Electronics and
Communication Engineering in GIET in Jawaharlal Nehru Technological from an unknown speaker will be identified to be one
University,Kakinada,INDIA in 2010,PH:9493493302. among the N enrolled speakers. Thus there is a risk
Email:[email protected]
*Dr V.Sailaja received Ph.d degree in Speech Processing from Andhra
of false identification. Therefore, closed set mode should
University in 2010, INDIA,PH:9491444434, be employed in applications where it is surely to be used
Email:[email protected] always by the set of enrolled speakers. On the other
*Mr Shuaib Akram studying IV B.Tech in Electronics and Communication
Engineering, Jawaharlal Nehru Technological University , hand, speaker identification system which is able to
Kakinada,INDIA,PH:9703976497, Email:[email protected] identify the speaker who may be from outside the set of
IJSER 2011
113

ISSN 2229-5518
N enrolled speakers is known as open-set speaker integration method to re-score the hypothesis scores,
identification. In this case, first the closed-set speaker measure the distance between them and combine them.
identification system identifies the speaker closest to the
test speech data. The speaker identification system is
divided into text independent speaker identification and 2. FINITE MULTIVARIATE GENERALIZED
text dependent speaker identification. Among these two, GAUSSIAN MIXTURE SPEAKER MODEL
Text Independent Speaker Identification is more The Mel frequency cepstral coefficients (MFCC) are used
complicated in open test. to represent the features for speaker identification. In
the set up used, the magnitude spectrum from a short
Speaker Identification: frame is processed using a mel-scale filter bank. The log
Given different speech inputs X1,X2,, Xc energy filter outputs are the cosine transformed to
simultaneously recorded through C multiple produce cepstral coefficients. The process is repeated
microphones, whoever has pronounced X1,X2,.Xc every frame resulting in a series of feature vectors [1].
among registered speakers S={1,2,..C} is identified by We assume that the Mel frequency cepstral coefficients
equation (1). Each speaker is modeled by GGMM k. of each are assumed to follow a Finite Multivariate
Generalized Gaussian Mixture Distribution. Therefore
the entire speech spectra of the each individual speaker
can be characterized as a M component Finite
= p(k | X1,X2,.Xc). multivariate Generalized Gaussian mixture distribution.
(k | , ,. ). ( ) The probability density function of the each individual

= (1)
( , , ) speaker speech spectra is
p(x /) = b (x /) (4)
By using Bayess rule equal prior probability (ie. where, x = x j=1,2,,D; i=1,2,3,M;
P(k.)=1/C), and the conditional independency between t=1,2,3,,T is a D dimensional random vector
different speech inputs X1,X2,.Xc given speaker model representing the MFCC vector. is the parametric set
p , and not in that p(X1,X2,..Xc) is the same for all such = (,, ) is the component weight such that
speakers, equation 1 can be simplified as, =1
b (x /) is the probability density of ith acoustic class
= ( | k) (2) represented by MFCC vectors of the speech data and
the D-dimensional Generalized Gaussian (GG)
distribution (M..Bicego et al (2008)) [5] and is of the form
Taking the logarithm of equation 2, we obtain [ ()] /
b (x (, , ) = [ ( ) ( , )]
exp (5)
= ( | k) (3) where,z() = and
( / )
The identity of the speaker can be A(, ) = ( / )
(6)
determined by the sum of hypothesis log likelihood
scores obtained from C microphones. In a distance and x = |x | stands for the
talking environment, however, the log likelihood score
itself in equation(3) is expected to degraded, ie. Its l norm of vector x, is a symmetric positive definite
reliability cannot be ensured. Furthermore, a variety of matrix. The parameter is the mean vector, the function
causes such as the location of the speaker or the noise, A () is a scaling factor which allows the var(x) = 2 and
the direction of the speaker, and the distance can have a is the shape parameter when =1, the Generalized
different effect on each microphone. Therefore the Gaussian corresponds to a laplacian or double
identification result obtained from a microphone can be exponential Distribution. When =2, the Generalized
better than the others. In such cases, the simple Gaussian corresponds to a Gaussian distribution. In
integration is greatly affected by the incorrect limiting case + Equation (5) Converges to a uniform
classification of single channel. Thus, we propose a new distribution in (-3, +3) and when o +, the
distribution becomes a degenerate one when x=. The
IJSER 2011
114

ISSN 2229-5518
generalized Gaussian distribution is symmetric with 3. EXPERIMENTAL SETUP

respect to , The proposed Generalized Gaussian Mixture Model in a
The variance of the variate x is multi microphone environment are evaluated with a
var(X) = (7) database uttered by 25 speakers. For each speaker, there
The model can have one covariance matrix per a are 10 conversational sentences, which are recorded in
Generalized Gaussian density of the acoustic class of single session. Each sample is of 2 seconds. Speech
each speaker. The covariance matrix can also be a full samples are recorded by using 8 micro phones. Each of
or diagonal. In this chapter the diagonal covariance which recorded at centre and diagonal. They were then
matrix is used for speaker model. This choice is based re-recorded again by playing them back with a loud
on the initial experimental results. As a result of speaker placed at each position, means 1m, 3m and 5m
diagonal covariance matrix for the feature vector, the by using 8 microphones which were placed at different
features are independent and the probability density locations. These are used to train GGM and to estimate
function of the feature vector is the distribution of average log likelihood estimation.
( , )
b (x |() = = f x
( , )
4. EXPERIMENTAL RESULTS:
(8)
The model parameters are estimated and initialized by
As the number of N-best hypotheses per channel (N- best
the EM algorithm with k-means [7].
classification results) employed for identification
increases, the performance of the proposed method
The updated equation for estimating the model outperform the earlier existing methods.
parameters are
Table1: Speaker identification accuracy
The updated equation for estimating i is
LOCAL 1m 3m 5m
() , ()
( )
= () (9)
, ()
CH C D C D C D
Where ( ) = ( ) , ()
are the estimates obtained at
the ith iteration. 0 94.9 95.3 68.3 67.8 75.7 74.6
The updated equation for estimating is

1 91.8 91.6 66.2 66.9 63.9 64.5
2 94.0 93.4 73.1 72.8 80.6 81.2
,
( ) ( , )( )
= ,
(10) 3 91.8 92.3 62.7 62.5 56.1 56.9
( , ( ))
Where, A N, is some function which must be equal 4 94.6 93.9 54.4 55.6 57.5 58.4
to unity for i = 2 and must be equal to for 1, in
5 94.4 93.8 69.7 70.3 56.5 57.4
the case of N=2, we have also observed that A N,
must be an increasing function of . 6 96.2 95.6 61.8 62.6 60.6 61.4
The updated equation for estimating is
7 94.0 93.5 57.5 58.9 62.2 63.1
( , ( )) ()

AVG 94.0 93.7 64.2 64.7 64.1 64.7
( )= ( , ( ))
(11)
LS 93.0 93.9 67.4 80.4 70.9 78.3

AD 93.0 93.5 68.3 80.9 74.3 81.3
The number of mixture components is initially taken for
K Means algorithm by drawing the histogram of the GGM 94.0 93.7 64.2 64.7 64.1 64.7
first Mel frequency cepstral coefficient of the speech
data.
IJSER 2011
115

ISSN 2229-5518
The figure represent the identification accuracies of the

identification methods, as the number of N- best
classification result increases to 25, which corresponds to
total number of registered speakers .Table.1 corresponds
to the performance of the identification methods in case
of N=25. In terms of the base line method (LS), if
we approximate k in Eq (3) to k N-best hypotheses, it is
possible not to obtain the log likelihood score of the
hypothesis from a certain channel depending on the
order of hypotheses. Thus we considered only when N is
equal to 25. Despite only using 2-best classification
results per channel, the proposed method (GGM) is
comparable to, or even better than, the baseline methods
using all possible hypotheses, which consist of all the
registered speakers. And the proposed identification
method can achieve its best performance with less than
Figure 1: Plot of DET Curves for a speaker
eight best hypotheses per channel. Table 2 shows the
identification evaluation.
average percentage of correct identification for various
speaker identification models.
CONCLUSION
In this paper we propose Text Independent Speaker
Table 2 Avg. percentage of correct identification vs. Identification model based on Finite Multivariate
speaker identification models. Generalized Gaussian Mixture Model, to improve
identification accuracy in a Multi-microphone
Speaker Model % of Accuracy environment. In a relatively close-talking environment
(1m), Generalized Gaussian model maintain high
LS 93.0 identification accuracy and they are superior to
identification result of the best channel. As the distance
AD 93.0 between the speaker and microphone increases, the
Generalized Gaussian Mixture Model shows more
GGM 94.0 reliable identification performance than other existing
models which are given in the table (1). For this, a text
independent speaker identification model is developed
*LS- Least Squared, AD-Adaptive Distribution, with the assumption that the feature vector associated
with speech spectra of each individual speaker follows a
GGM- Generalized Gaussian Model Finite Multivariate Generalized Gaussian Mixture Model.
A graph between False Alarm probability and Miss- The model parameters are estimated and initialized by
probability called as DET curve is drawn below. If false using EM algorithms with K-means. An experimentation
alarm probability and Miss-probability are equal, the with 25 speakers speech data revealed that this Text
graph can be approximated as hyperbola. Independent Speaker Identification using Finite
Multivariate Generalized Gaussian Mixture Model
outperform the earlier existing Text Independent models.
IJSER 2011
116

ISSN 2229-5518
References:
[1] B. Narayana Swamy and R. Gangadharaiah Extracting

Additional Information From GAUSSIAN MIXTURE MODEL
PROBABILITIES for improved Text Independent Speaker
Identification. Proc. ICASSP, vol.1, pp.621-624, Philadelphia,
USA, March 2005.
[2] D.A.Reynolds, R.C.Rose. E.M. Hofstetter Integrated Models
of signal and background with application to speaker identification
in noise IEEE Trans. On speech and audio Processing, vol 2. No.
2. April 1994.
[3] D.A.Reynolds, R.C.Rose. Robust Text Independent Speaker
Identification using Gaussian Mixture Models IEEE Trans. On
speech and signal Processing, 3(1), January 1995, 72-83.
[4]. I.A. McCowan, J. Pelecanos and S. Sridharan, Robust speaker
recognition using microphone arrays Proc. Speaker Odyssey,
pp.101-106, Crete, Greece, June 2001.
[5] Md M. Bicego, D Gonzalez, E Grosso and Alba Castro (2008)
Generalized Gaussian distribution for sequential Data
Classification IEEE Trans. 978-1-4244-2175-6.
[6] Q. Lin, E. Jan, and J. Flanagan, Microphone arrays and
speaker identification, IEEE Trans. Speech Audio Process, vol.2,
no.4, pp.622-628, Oct. 1994
[7] Sailaja, K Srinivasa Rao & K V V S Reddy (2010) Text
Independent Speaker identification with Multi Variate Gaussian
Mixture Model International Journal of Information Technology
and Knowledge Management July-December 2010, Volume 2, No.
2, pp. 475-480.
IJSER 2011
117
ISSN 2229-5518
Novel Object Removal in Video Using Patch

Sparsity
B. Vidhya, S. Valarmathy
Abstract The process of repairing the damaged area or to remove the specific areas in a video is known as video inpainting.
To deal with this kind of problems, not only a robust image inpainting algorithm is used, but also a technique of structure
generation is used to fill-in the missing parts of a video sequence taken from a static camera. Most of the automatic techniques
of video inpainting are computationally intensive and unable to repair large holes. To overcome this problem, inpainting method
is extended by incorporating the sparsity of natural image patches in the spatio-temporal domain is proposed in this paper. First,
the video is converted into individual image frames. Second, the edges of the object to be removed are identified by the SOBEL
edge detection method. Third, the inpainting procedure is performed separately for each time frame of the images. Next, the
inpainted image frames are displayed in a sequence, to appear as a inpainted video. For each image frame, the confidence of a
patch located at the image structure (e.g., the corner or edge) is measured by the sparseness of its nonzero similarities to the
neighboring patches to calculate the patch structure sparsity. The patch with larger structure sparsity is assigned higher priority
for further inpainting. The patch to be inpainted is represented by the sparse linear combination of candidate patches. Patch
propagation is performed automatically by the algorithm by inwardly propagating the image patches from the source region into
the interior of the target region by means of patch by patch. Compared to other methods of inpainting, a better discrimination of
texture and structure is obtained by the structure sparsity and also sharp inpainted regions are obtained by the patch sparse
representation. This work can be extended to wide areas of applications, including video special effects and restoration and
enhancement of damaged videos.
Index Terms Candidate patches, edge detection, inpainting, linear sparse representation, patch sparsity, patch propagation,
texture synthesis.
1 INTRODUCTION
V ideo inpainting is the process of filling - in of the

missing regions in a video. Inpainting technique is
the modification of the images in an undetectable
inpainting methods using texture synthesis is in the size
and characteristics of the region to be inpainted. For
texture synthesis, the region can be much larger with the
form. Nowadays video has become an important media main focus being the filling in of two-dimensional
of communication in the world. Image inpainting is repeating patterns that have some associated stochasticity
performed in the spatial domain where as video i.e. textures. In contrast, inpainting algorithms
inpainting is performed in the spatio-temporal domain. concentrate on the filling in of much smaller regions that
Video inpainting also plays a vital role in the field of are characterised by linear structures such as lines and
image processing and computer vision as that of the object contours. Criminisi et al. [3] presented a single
image inpainting. There are numerous goals and algorithm to work with both of these cases provided both
applications of the video inpainting technique from the textures and structures are present in images.
restoration of damaged videos and paintings to the Nowadays, various researches are performed in the
removal or replacement of the selected objects in the field of video inpainting due to the varied and important
video. applications of an automatic means of video inpainting.
Video inpainting is used to remove objects or restore The main applications include undesired object removal
missing or tainted regions present in a video by utilizing such as removing unavoidable objects like birds or
spatial and temporal information from neighbouring aeroplane that appear during filming, to censor an
scenes. The overriding objective is to generate an obscene gesture or action that is not deemed appropriate
inpainted area that is merged seamlessly into the video so for the target audience, but for which it would be
that visual coherence is maintained throughout and no infeasible or expensive to re-shoot the scene.
distortion in the affected area is observable to the Another main application is the restoration of the
human eye when the video is played as a sequence. videos that are damaged by scratches or dust spots or
Video is considered to be the display of sequence of frames that may be corrupted or missing. When the video
framed images. Normally twenty five frames per second is transmitted over unreliable networks, there is a
are considered as a video. Less than twenty five frames possibility of losing significant portions of video frames.
per second will not be considered as a video since the To view the video again in its original form, it is
display of those will appear as a flash of still image for the necessary to repair these damaged scenes in a manner
human eye. that is visually coherent to the viewer.
The main difference between the video and image Initially, all the above applications were performed
IJSER 2011
118
ISSN 2229-5518
manually by the restoration professionals which were region for every time frame of image to reproduce the
painstaking, slow and also very expensive. Therefore original image. Last, the inpainted image frames are
automatic video restoration certainly attracted both displayed to form the inpainted video. Here a video of
commercial organizations (such as broadcasters and film short duration is considered for inpainting and the
studios) and private individuals that wish to edit and temporal domain information for each image frame is
maintain the quality of their video collection. utilized to display the inpainted image frames as a video.
Bertalmio et al. [1], [2] designed frame-by-frame PDEs In this paper, Section 2 gives an overview of the
based video inpainting which laid platform for all image inpainting using extended examplar-based
researches in the field of video inpainting. Partial inpainting method. In Section 3, the method of video
Differential Equation (PDE) based methods are mainly inpainting is defined. The experiments and the results are
edge-continuing methods. In [2], the PDE is applied discussed in the Section 4. Finally, conclusion and the
spatially, and the video inpainting is completed future research of the work are discussed in Section 5.
frame-by- frame. The temporal information of the video is
not considered in the inpainting process.
Wexler et al [6] proposed a method for space-time
2 IMAGE INPAINTING
completion of large damaged areas in a video sequence. The most fundamental method of image inpainting is the
Here the authors performed the inpainting problem by diffusion based image inpainting method in which the
sampling a set of spatial-temporal patches (a set of pixels unknown region is filled by diffusing the pixel values of
at frame t) from other frames to fill in the missing data. the image from the known region. Another method of
Global consistency is enforced for all patches surrounding image inpainting is the examplar based image inpainting
the missing data so as to ensure coherence of all in which the region is filled by propagating the patch
surrounding space time patches. This avoids artefacts values of the known region to unknown region. In the
such as multiple recovery of the same background object previous work of this paper, an examplar based image
and the production of inconsistent object trajectories. This inpainting was proposed by incorporating the sparsity of
method provides decent results, however it suffers from a natural image patches.
high computational load and requires a long video
sequence of similar scenes to increase the probability of
correct matches. The results shown are of very low Source region p
resolution videos, and the inpainted static background
was different from one frame to another creating a ghost
effect. Significant over-smoothing is observed as well.
Video inpainting meant for repairing damaged video
was analysed in [4], [7] which involves gamut of different
techniques which made the process very complicated. p
Target region
These works combine motion layer estimation and
segmentation with warping and region filling-in. We seek
a simpler more fundamental approach to the problem of a b
video inpainting.
Inpainting for stationary background and moving Fig.1 Patch selection
foreground in videos was suggested by Patwardhan et al.
[5]. To inpaint the stationary background, a relatively Fig.1 (a) shows that is the missing region , is the known region,
simple spatio-temporal priority scheme was employed and is the fill-front of patch propagation (b) shows the two
where undamaged pixels were copied from frames examples of the surrounding patch p and p which are located at
temporally close to the damaged frame, followed by a edge and flat texture region respectively
spatial filling in step which replaces the damaged region
with a best matching patch so as to maintain a consistent The process of filling in of the missing region using the
background throughout the sequence. Zhang et al., [7] image information from the known region is called as the
proposed a motion layer based object removal in videos image inpainting. Let I be the given image with the
with few illustrations. missing region or target region . In the examplar based
In this paper, video inpainting for static camera with a image inpainting, the boundary of the missing region is
stationary background and moving foreground is also called as the fill front and is denoted by . A patch
considered in the spatial temporal domain. First, the centered at a pixel p is denoted by p.
video is converted into image frames. Second, the edges The examplar based image inpainting is based on the
are found by using SOBEL edge detection method. Next, patch propagation. It is done automatically by the
the object to be removed is inpainted using novel algorithm by inwardly propagating the image patches
examplar based image inpainting using patch sparsity. from the source region into the interior of the target
The known patch values are propagated into the missing region patch by patch. Patch selection and patch
IJSER 2011
119
ISSN 2229-5518
inpainting are the two basic steps of patch propagation video inpainting is performed in both spatial and
which is to be iterated continuously till the inpainting temporal domain. It is used to remove objects or restore
process is complete. missing regions in the video sequence. Video inpainting
Fig.1 shows the patch selection process which is used to may also be considered as the combination of frame by
select the patch with highest priority for further frame image inpainting.
inpainting. The sparseness of the nonzero similarities of a
patch to its neighboring patches is called as the structure
sparsity which is used for assigning patch priority. As Input video
shown in the Fig. 1(b), p and p are the patches which
are centered at pixel p and p which lie in the edge
structure and the flat texture region of the image
respectively. The patch p has sparser nonzero
similarities than patch p, so it is given larger patch Framed Images
priority. The patch inpainting step is performed.
Edges of the Target object

q
q Compute the similarities of patch with its

neighboring patches in the source region

p p
Compute the patch priority
a b
Inpaint the selected patch

Fig.2 Patch inpainting
Fig.2 (a) shows that for the selected patch p , sparse linear
combination of candidate patches { q, q... qN} is used to infer
the missing pixels in patch p . (b) Shows the best matching patch
in the candidates set has been copied into the position occupied by
Is the filling of
p , thus achieving partial filling of .
the missing region
is same as the
Fig.2 shows the procedure of patch inpainting which is No
known region?
used to selecte the patch on the boundary for inpainting.
The selected patch on the fill-front is the sparse linear
combination of the patches in the source region
Yes
regularized by sparseness prior. In this paper, a single
best match examplar or a certain number of examplars in
Exit
the known region to infer the missing patch is not used.
The most likely candidate matches for p lie along the
boundary between the two textures in the source region, Iterate the above process for all image frames
e.g., q and p. The best matching patch in the
candidates set has been copied into the position occupied
by p, thus achieving partial filling of . The target
region has, now, shrunk and its front has assumed a Display all the individual inpainted
different shape. Thus image inpainting is performed by image frames in order
the sparse linear combination of candidate patches
weighted by coefficients, in which only very sparse
nonzero elements exist. Inpainted video as output
Fig. 3 Flow Diagram of the Video Inpainting

3 VIDEO INPAINTING METHOD
Video is the display of the image frames in sequence. Generally all the natural images are composed of
Image inpainting is done in spatial domain, whereas the structures and textures. The primal sketches of an image
IJSER 2011
120
ISSN 2229-5518
like the edges, corners, etc are referred to as the structure which has more sparsely distributed nonzero similarities
of an image and the image regions with feature statistics is prone to be located at structure due to the high
or homogenous patterns including the flat patterns are sparseness of structures. Confidence of structure for the
referred to as the texture of an image. patch and the data term of the structure are illustrated in
Patch priority must be defined in such a way that it the Fig 4. (c) and (d) respectively.
should be able to differentiate the structures and textures
of an image. The structure sparsity is used for assigning
priority to the patches to be inpainted. It can also be
defined to measure the confidence of a patch located at
structure instead of texture.
The structures are sparsely distributed in the image
domain. The neighboring patches of a particular patch
with larger similarities may also be distributed in the
same structure or texture as the patch of interest to be
inpainted. To overcome such cases, the confidence of the
structure can be modelled for a patch by measuring the a b
sparseness of its nonzero similarities to the neighboring
patches.
The input to the object removal algorithm is the framed
images. Initially, the structure and texture values of the
given image are calculated and the object to be removed
is found. Edges of the object to be removed are obtained
through the edge detection and the similarities of the
patch with its neighboring patches in the source region
are computed.
Then, the patch priority is obtained by multiplying the
c d
transformed structure sparsity term with patch
confidence term. The patch with highest priority is
Fig.4 Important terms in the object removal method
selected for further inpainting. The above process is
repeated until the missing region is completely filled by
Then the patch priority is computed by the product of
the known values of the neighboring patches.
the data term and the confidence term. The patch p with
The sequence of display of the individual time frame of
the highest patch priority on the fill-front is selected.
inpainted image is the video inpainting. Image inpainting
is performed for each image frame, and all the inpainted
image frames are added to form the video. Fig. 3
illustrates the overall flow diagram of the video
inpainting process.
4 RESULTS AND DISCUSSION

Fig 4 illustrates the important terms used in the
inpainting algorithm. Fig 4. (a) shows the original image a b c
I which is given as input to the novel examplar-based
image inpainting method. The natural images are
generally composed of structures and textures.
A good definition of patch priority should be able to
better distinguish the structures and textures, and also be
robust to the orientation of the fill-front. In this paper, a
novel definition of patch priority is proposed to meet
these requirements. The missing region or the target
region is first analysed using the algorithm as shown in d e f
the Fig 4. (b). Fig.5 Steps in Video Inpainting
The structure sparsity is defined to measure the Fig. 5.(a),(b) and (c) shows randomly selected image frame of the
confidence of a patch located at structure instead of original video sequence. Fig 5 (d), (e) and (f) shows edge detection
texture. For a certain patch, its neighboring patches with of the image frames (a), (b) and (c) respectively.
larger similarities are also distributed in the same
structure or texture as the patch of interest. The patch Then inpainting is performed based on the patch
IJSER 2011
121
ISSN 2229-5518
propagation by inwardly propagating the image patches frames of each time frame is displayed as the inpainted
from the source region into the interior of the target video. Experiments showed that the proposed
region patch by patch. The above procedure is repeated examplar-based patch propagation algorithm can
for each image frames and displayed in the sequence as produce sharp inpainting results consistent with the
video. surrounding textures.
In this paper, static camera with constant background is
considered. In the future, background with multiple
scales and orientations, moving camera and high
resolution videos will also be investigated. Video
inpainting for long run (time duration) video shall be
accomplished. This work can be extended to wide areas
of applications, including video special effects and
restoration and enhancement of damaged videos.
a b c REFERENCES
[1] M. Bertalmio, G. Sapiro, V. Caselles, and C. Ballester, Image
Fig. 6 Inpainted Image frames of the video sequence inpainting, in Proc. SIGGRAPH, 2000, pp. 417424.
[2] M. Bertalmio, A. L. Bertozzi, and G. Sapiro, NavierStrokes,
Fig 6 (a), (b) and (c) illustrates the inpainted image frames for the fluid dynamics, and image and video inpainting, in Proc. IEEE
corresponding Image Frames shown in Fig 5 (a), (b) and (c) Computer Society Conf. Computer Vision and Pattern Recognition,
respectively for the given video sequence. pp. 417424, 2001.
[3] Criminisi, P. Perez, and K. Toyama, Region filling and object
Fig. 5 (a), (b) and (c) illustrates the randomly selected removal by examplar-based image inpainting, IEEE Trans. Im-
image frames in a video. Fig. 5 (d), (e) and (f) shows the age Process., vol. 13, pp. 12001212, 2004.
edge detected image frames of Fig. 5 (a), (b) and (c) [4] J. Jia and C. K. Tang, Image repairing: Robust image synthesis
respectively. The selected patch p on the edge is by adaptive and tensor voting, in Proc. IEEE Computer Society
inpainted by the corresponding pixels in the sparse linear Conf. Computer Vision and Pattern Recogition, pp. 643650, 2003.
combination of examplars to infer the patch in a frame [5] K.A. Patwardhan, G. Sapiro, and M. Bertalmio, Video inpaint-
ing of occluding and occluded objects", in Proc. ICIP 2005.Vol.
work of sparse representation. The fill-front and the
II, pp. 69-72.
missing region are updated instantly. For each
[6] Y. Wexler, E. Shechtman, and M. Irani, Space-time video com-
newly-apparent pixel on the fill-front, its patch
pletion, Proceedings. 2004 IEEE Computer Society Conference on
similarities are computed with the neighboring patches Computer Vision and Pattern Recognition, vol. 1, 2004.
and its patch priority. [7] Y. Zhang, J. Xiao, and M. Shah, Motion layer based object
The selected patch is used for inpainting. By iterating removal in videos, 2005 Workshop on Applications of Computer
the above process, the filling region isinpainted Vision, 2005.
successfully for the entire image frames. Fig. 6(a), (b) and
(c) illustrates the inpainted image frames of the original
image frames of athe video as shown in Fig. 5 (a), (b) and
(c) respectively. The inpainted image frames are added
together to display as a video sequence.
5 CONCLUSION
Novel patch propagation based inpainting algorithm for
video inpainting is proposed in this paper. It is mainly
focussed on the object removal in the image frames. Patch
priority and patch representation are the two major steps
involved in the proposed examplar-based inpainting
algorithm for an image. Structure sparsity was
represented by the sparseness of the patch similarities in
the local neighborhood. The patch at the structure with
larger structure sparsity is given higher priority and is
used for further inpainting. The sparsest linear
combination of candidate patches under the local
consistency was synthesized by the patch sparse
representation. Video is represented by the display of the
sequence of the image frames. Hence the inpainted image
IJSER 2011
122

ISSN 2229-5518
New Formula of Nuclear Force

Md. Kamal Uddin
Abstract -It is well established that the forces between nucleons are transmitted by meson. The quantitative explanation
of nuclear forces in terms of meson theory was extremely tentative & in complete but this theory supplies a valuable point
of view . it is fairly certain now that the nucleons within nuclear matter are in a state made rather different from their free
condition by the proximity of other nucleons charge independence of nuclear forces demand the existence of neutral
meson as amongst the same type of nucleolus (P-P) or (N-N). this force demand the same spin & orbital angular
momentum. The exchange interaction in produced by only a neutral meson. The involving mesons without electric charge,
that it gives exchanges forces between proton & Neutron & also therefore maintains charge in dependence character. It is
evident for the nature of the products that neutral mesons decay by strong & weak interaction both. It means that neutral
mesons constituents responsible for the electromagnetic interaction. Dramatically neutral mesons plays important role for
electromagnetic & nuclear force both.
Index Terms - Restmass energy,Mesons,photons,protons,neutrons,velocity of light,Differentiation
1. INTRODUCTION
IT is well established that the forces

between nucleons are transmitted by
application of quantum mechanics to
nuclear phenomena gives us confidence in
meson. The quantitative explanation of general use of quantum mechanics for the
nuclear forces in terms of meson theory description of the force between heavy
was extremely tentative & incomplete, but particles in nuclei.
this theory supplies a valuable point of Where dm0c2 = either rest mass energy of 0
view. Yukawa first pointed out that mesons(For neutral theory),or rest mass
nuclear force can be explained by assuming energy of +, -& 0mesons( for symmetrical theory)
that particle of mass about 200 times the dm0 =either mass of 0 mesons or mass of
electron mass(mesons) exist & can be + , -& 0 mesons
emitted & absorbed by nuclear m0 = mass of nucleons
particles(neutrons & protons) with such an m0cdc = rest mass energy of nucleons
assumption a force between nuclear dr= Range of nuclear force, which can be
particles of right range & right calculated from differentiation of Nuclear
shape(rapid decrease at large distances is radius.( The force between two nucleons is
now obtaining. attractive for distance r(radius) greater
Now we have the rest mass energy = m0 c2 than dr (range) & is repulsive
Differentiating with respect to r (Inner radius otherwise).This strongly suggests & well
at which nuclear force comes into play) proved that to some degree of
d(m0c2) approximation the total isotopic spin T is a
dr constant of the motion & is conserved in
= c2dm0 + m0d(c2) = c2dm0 + all processes, at least with a high
dr dr dr probability.
mo d(c2) dc = c2dm0 + 2m0c.dc dc= The average velocity of neutron &
dc dr dr dr proton. A large velocity is used in nuclear
. This force is short range, attractive & along disintegration.
the line joining the two particles (central
force).(The wide success of this first c = Velocity of light
IJSER 2011
123

ISSN 2229-5518
2 = multiplicity of interacting particles interact only with neutral mesons(

particles is given by (2T+1), the isotopic spin neutral theory) or(2) that they interact
has no such meaning for leptons or a gamma equally strongly with neutral, positive &
rays negative mesons(symmetrical theory). It is
1 = either multiplicity of obvious that the part of the force which
0 mesons or +, - & 0 mesons (evidence of does not depend on the spin of the
involving of mesons (all type)) nuclear does not fulfill any useful
Where T = Vector sum of function in the theory. The force between
isotopic spin of proton & proton, neutron& proton & neutron are result from the
neutron, neutron & proton The success of transfer of positive meson from the former
these applications supplies additional to the latter or a negative meson in the
support for the hypothesis of the charge opposite direction. So there vector sum of
independence & charge symmetry of component of isotopic spin of these particle
nuclear force. As the nuclear interactions do must be zero.
not extend to very large distances beyond The charges on charged
the nuclear radius & this character is mesons must be equal in their magnitude.
useful to solve the problem. The Full
charge independence for any system in
which the number of neutrons equals the Charge independence of nuclear forces
numbers of protons, this formula give the demand the existence of 0 meson as amongst
evidence the charge symmetry, merely the same type of nucleons (p-p) or (N-N).This
means that the neutron-neutron & proton force demand the same spin & orbital
proton interaction are equal but says angular momentum. Positive pions are not
nothing about the relations of neutron able to surmount the nuclear coulomb
proton interaction to others. Nuclear forces barrier & there fore undergo spontaneous
are symmetrical in neutrons & protons. i.e. decay while negative pions are captured
the force between two protons are the by nuclei. The exchange of a pion is thus
same as those between two neutrons. This equivalent to charge exchange. we can think
identity refers to the magnitude as well as of nucleons as exchanging their space &
the spin dependence of the forces. spin co-ordinates In the neutral theory,
therefore neutron & protons are completely
Now we can see the following reaction equivalent & indistinguishable as far as the
P+ P ----- P+P+0 associated meson fields are concerned. Such
P+N -----P+N+ + particle decay into two gamma rays. These
N +P----- N+ P + - gamma rays are 0 rest systems are emitted
These reactions are soon as Y + P -- in opposite direction & therefore spin 0 must
P + 0 be Zero as the spin of photon is unity. It is
The capture of photons can effect evident from the nature of the products
the production of mesons by an that neutral mesons decay by the
electromagnetic interaction, decay electromagnetic interaction while charged
electromagnetically since these processes pions decay by strong & weak interaction
involve no change of strangeness. both. It means that neutral mesons
P + P --- P + P + 0 constituents responsible for the
P + N ----- P + N + 0 electromagnetic interaction. We know that
It is found that only two assumptions neutron & proton can change into one
are in agreement with theoretical & another by meson capture. Protons &neutron
experimental facts, notably the equality of can transform into each other by capture of
the forces between two like & two unlike positive & negative pion respectively, or get
nuclear particles in the singlet state. These transform into the same particle through
assumption are either(1) that nuclear neutral meson interaction. During these
IJSER 2011
124

ISSN 2229-5518
transformation either an emission or an stable than hydrogen atom.Becouse binding

absorption of meson is essential. The energy of helium atom is larger than that
attraction between any nucleons can arise of hydrogen atom, so in this process large
from the transfer of a neutral meson from energy is released.( here the negative pion
one nucleon to the other. If the meson has significant role to produce nuclear
were assumed to be charged ( positive or interaction. While charge pion main tain the
negative) the resulting force between character of proton.) Now, as the hydrogen
nuclear particles turned out to be of the nucleus converted into helium nucleus,
exchange type which had been successful These are happen when resonances of
in the interpretation in nuclear physics. The nucleons is in excited state. The scattering
mesons must obey Bose statistics because cross- section can be interpreted by
they are emitted in the transformation of a assuming that in the strong interactions the
neutron into a proton( or vise versa) both total isospin is converted as well as the
obey Fermi statistics. third component. The total isotopic spin of
the system is 3/2.The ratio of radius &
Important point-- range is about 2:1. It shows that those
1. The deuteron does posses nuclei has maximum number of nucleons
measurable properties which might serve as are most stable than less mass number
a guide in the search for the correct nuclei.
nuclear interaction. The mass number of
deuteron A is very minimum. From the 2*.Also, Strongly, The
findings we must regard the deuteron as velocity of light depends
loosely bound. The deuteron consist of two upon range of the nuclear
particle roughly equal mass M, so that the force. The velocity of light
reduced mass of the system is 1/2M.The equal like photons, lack
deuteron has spin T=1, the neutron & mass & force carrying
protons spins might be a parallel particle of other forces like
combination. The magnetic moment of strong force. Because range
deuteron will, therefore be sum of is variable, then the velocity
magnetic moment of proton & neutron. of light must be variable.
According to conclusion, As the range As, velocity of light= Range
becomes larger in deuteron nucleus & they of nuclear force(distance
becomes more unstable. travel by meson)/ Life time
*When light nuclei of hydrogen atom of resonances. In this
comes within the range of nuclear force relation we can see that the
they can fuse together to form helium velocity of light must be
nucleus. In this process (fusion) the range is variable. It is clear that
not effected. However the force is twice of the fundamental particles
the hydrogens nuclear force & so on. are not wholly
Further the energy requires to bring independent, The neutron
nucleons inside the range of force is twice observed to change
of the rest energy of hydrogen nucleus. In spontaneously into a
other word, we can say that the minimum proton. Neutron decay
energy requires to form a helium nucleus is takes on the average some
twice of the hydrogen rest mass energy. thousands seconds for free
The mass of He atom (alpha-particle) equal neutron, whereas within a
to the four times of the hydrogen atom, so nucleus the characteristic
that the nuclear force of helium nucleus is time between nucleon-
four times stronger than that of hydrogen nucleon collisions is 10-24
atom. It means that the He atom is more seconds. For a satisfactory
IJSER 2011
125

ISSN 2229-5518
picture it is often enough to depths are not identical but depends upon
think of the nucleus as a the state of their actual binding in the
grouping of protons & potential well. The range also depends
neutrons interaction, with upon mass number A & binding energy.
the appearance or we know that the atomic mass number A
disappearance of is approximately equal to a twice the
photons. One atomic number Z. For the light &
should be noted that this intermediate nuclei. It shows that light
relation hold only inside nuclei prefer to add nucleons is n-p pair.
the nucleus. Out side the i.e there is a strong interaction between
nucleus the evident is to neutrons & protons. The range of nuclear
be contrary. It is a fact no force depends on the mass number A &
body(even mesons or the velocity of light depends on the range,
gamma rays ) can have so it is obviously thought that the spin &
velocity greater than the velocity of light depends on the mass
velocity of light. From this number A of the nucleus & spin is zero
formula we can find the or an integer for A even & is an odd half
nuclear force acts between integral for A odd. The total rest mass
the pair of nucleons & energy also depends on the mass number
does not influenced by the A. For increasing of rest energy, we must
presence of neighboring increase the mass number A. obviously, the
nucleons. It is necessary rest mass energy must be depends on the
that any one particle must radial distance.
brings the velocity of light. . This is purely a quantum mechanical
We know that the nuclear effect. If the mass number A increases the
force is short ranged. Out range decreases, & the force are stronger.
side of the range it is This binding energy displays saturation
repulsive. effect. This property of the nuclear force
can be explain in term of exchange nature
*Range of nuclear force :- To show that the of nuclear force. It should be noted that
range of force is related to the mass of nucleons attract each other strongly only if
exchanged particle, assume that the they are in same orbital state. This formula
0-meson is contained virtually in a proton. prove the pauli hypothesis. This formula
If this virtual particle travels with the usually attributed to the effect of higher-
velocity light as might be expected for a order interactions in which two or more
field particle, then greatest distance the mesons are simultaneously transmitted
meson could travel in this time also known between the nucleons.
as range of the pion exchange force.
3. It would seem that in a nucleus 4.The velocity of light depends on the
consisting of the many nucleons the binding wavelength of it constituents, If the
energy per nucleon should increase with particles has longer wavelength then the
the increase of the mass number A . In range decreases & therefore force is
reality evidence is to contrary, the binding stronger. we can find the effective range of
energy per nucleons decreases with nuclear force in terms of the Compton wave
increasing mass number A, The binding length of pimeson. We know that
energies of the different nucleon placed at different(variable) constituents (color
various depths are not identical but depend particles) has different wavelength, so it is
upon the states of their actual binding in obviously thought the velocity light must
the potential well. The binding energies of be variable. Evidently, if meson interacted
the different nucleon placed at various with nucleons strongly enough to be
IJSER 2011
126

ISSN 2229-5518
responsible for the nuclear forces. Its mean meson comes to close together, the can
free path within the nucleus should be neutralize each other then the force
about the same as is that of a nucleons. between neutron & proton come into play.
Then one question must be arise, How & So obviously we can say that only neutral
why the velocity of light vary in free path meson plays important role in charge
or in vacuum. So far this problem is independent nuclear force. The
concern, the velocity of light influenced by mesons(positive & negative) can be
its internal matter, which must have absorbed by the nucleus of an element or
different value. However it may be it may be combine with the another meson
possible that the velocity of light equal at then the sum of the masses of these
all ranges within the nucleus. . The forces mesons converted into energy. This process
responsible for binding the individual is called annihilation of matter. Before this
particle inside the nucleus must therefore process one positive meson & one negative
be exceptionally strong.If the particles has meson unite to make neutral particles
motion then the material body has physical called k0 meson. The process of construction
significance otherwise not. It means the & destruction has proved very help full is
force between elementary particles depends considering the origin of universe. The
on the velocity of the body as well as neutral K-- mesons is a stable particle but
mass of body. It should be remarked that stability last for a small time. Its half life
the particles travels with velocity of light is of the order of micro seconds. This
are not a conservable quantities. In particle is an essential constituent of the
quantum theory, every field must be nucleons of all elements.We know that
quantized. These quanta produce a field, neutralmeson decay into two photons &
which is responsible for different forces. never into three photons. It is clear that
the neutral pions has been produced by
bombarding hydrogen & deuteron with
high energy photons. Gamma rays has
sufficient energy to maintain energy of
5.The emission of a charged meson will be nucleons then the nucleons produce the
accompanied by a change of charge of the neutral mesons.One can speak of the meson
emitting nuclear particle, Thus a neutron field associated with a proton(or a netron)
can only emits a negative meson or absorb because the nature (charge) of the nuclear
a positive meson & will thereby be particle does not change by emitting or
transformed into proton. When we consider absorbing a neutral meson.It has developed
the emission of one meson by a nuclear a theory of nuclear forces in which neutral
particle & reabsorbtion by another.It is way the equality of the forces between like
obvious that in this way no force will be & unlike nuclear particles.The theory
obtain between two nuclear particle of involving charged meson only giving no
same kind i.e. two neutrons or two forces between two like nuclear particles. It
protons. For the same kind particles, there is obvious that the negative meson &
is neutral meson is responsible for positive meson gave the symmetrical force
interaction.This solution would make the between protons & neutrons & the interact
interaction coused by neutral meson alone, equally strongly with these meson. An
Since for unlike particles the charged alternative way of explaining this equality
mesons given an additional contribution, is to assume interaction with neutral meson
while for like particles they do not, the only. Then the charge of the nuclear
total interaction will not be the same3 for particles( whether it is a proton or a
like & unlike particles in S-first state.This neutron) becomes entirely irrelevant & the
will lead to forces between a neutron & a equality of forces follows immediately. This
proton.The negative & positive charge alternative is discussed in present paper.
IJSER 2011
127

ISSN 2229-5518
6.According to the pauli principle only of neutral mesons move only a few atomic
two neutrons & two protons will found in diameters before they decay( so that it
the same orbital state. Therefore it is influenced few neighbors nucleons)& thus
possible to find four nucleons strongly are not affected by the matter through
bound or Alpha particle structure, also which they pass & thus nuclear force work
confirm by binding energy curve. The properly. It should be also noted that in whole
extraordinary stability of the alpha shows universe there is only mass will be conserved
that the most stable nuclei are those in and energy will be destroyed, then the mass
which number of nucleons & photons are will not change into energy.
equal. We can find it from this formula. It *It is enough to think that - mesons which
is obviously thought that the full charge form a nuclear cloud around the individual
independence for any system in which the
number of neutrons equal to the number nucleons & are in a virtual state get their
of protons. From conclusion, we get, requisite rest mass energy from the
Number of photons = number of nucleons= incident particle & are released from the
2(number of neutral mesons).The discovery nuclear binding potential. Nuclear binding
of the neutral meson & the fact that charge potential compensate the rest mass energy.
independence is now consistent with all It produce enough energy to maintain the
nuclear data, confirm fully the use of the rest mass energy for production of mesons.
symmetric meson theory, containing Since the rest mass energy of - mesons
positive, negative &neutral mesons about 275 Me, the threshold energy for a
described by three way wave functions. gamma rays to produce the rest mass
With the form of Yukawa potential for energy of these particles should be high.
scalar mesons, it is easy to see that the pi- But, if protons projectiles are used to
meson can not be scalar. This theory prove produce mesons, it requires a large
this argument. threshold as a particle with mass retains
some energy in the collision. It should be
Change of Law-- remarked that the binding potential is
Since there is no requirement for the independent of spin & range of the
conservation of Pions so there is no particle, when they compensate the rest
conservation law in rest mass energy & energy. The energy require to pull out the
even in the universe. This formula show that nucleons from the nucleus is less than half
there is no meaning of the word constant. of rest mass energy. The slow motion
There is no conservation law controlling neutron play this role. Similarly if the
the total number of Kaons or meson. The nucleus brings (from binding potential)
energy of formation of mesons comes from sufficient energy for the existing of nuclear
binding potential ( which has the energy to force. It maintains stability. In order to
formation of meson for a long time), but approach particle to within short range or
when these potential has not enough closer the energy of the approaching
energy, the production of pions end & particle should be very high.
nuclear force does not exist.
*The life time of any radioactive substances Since there is no

depends on the total number of pions limitation of formation of
production and another particles production. Mesons even in strong
pions are commonly formed in the decay interaction. This is due to
of koans, hyperons & resonant states. It high energy photons ( -
should be noted that pions are formed only at rays ) then this cyclic chain
high energy.Becouse of their short life time should be possible
IJSER 2011
128

ISSN 2229-5518
0 rays
+d d+ 0
This reaction shows that the kinetic
energy as well as potential energy of
nucleon in the nucleus will be over
and above of the rest mass energy. In
these phenomena the total charge of
fundamental particles are conserved.
It is reasonable to assume
that the nuclear force
between two protons has
the same characteristic as
that between neutron &
proton. The argument
about short range forces
involves both proton-
proton & neutron -
proton forces. The main
difference between proton
& neutron seems to be
the electric charge, & the
nuclear force apparently
does not arise from
charge. We assume there
fore that the potential
between two protons is
confined within some
short range as before,
although the value of
range need not
necessarily be the same.
Finally-- some of the peculiarities of

nuclear forces can enumerated as
follows
(a) short range character
(b) Large strength, the nuclear
potential energies are quite large.
(c) Exchange character & saturable
nature.
(d) Dependence on spin
(e) charge independence& charge
symmetry.
References1. Elementry nuclear

theory.H.A.Bathe
2.Nuclear physics, Roy &
Nigam
3.Nuclear physics, Srivastva
IJSER 2011
129
ISSN 2229-5518
Impact Fatigue Behaviour of fully dense

Alumina Ceramics with Different Grain Sizes
Manoj Kumar Barai, Jagabandhu Shit, Abhijit Chanda, Manoj Kr Mitra
Abstract: Impact fatigue behavior of fully dense alumina as we studied in this work. The effect of grain size on the impact
fatigue characteristics has been found out using a simple impact fatigue test set-up. Low grained alumina was prepared using
optimized slip mcasting technique where as higher grained samples were made following conventional powdered metallurgy
technique (Compaction through isostatic pressing and subsequent solid sate sintering. All though the mechanical behavior
(e.g.hardness, toughness) was better in fine grained alumina, It was more succeptable to impact fatigue. Some of the sampled
in higher grained alumina samples were little bit elongated in shape with aspect ratio close to 2-2.5. Fractoghaphy revealed that
crack propagation was predominantly mixed mode. The elongated grains promoted bridging across crack front and caused
higher resistance to fatigue.
Index Terms dynamic, element, factor, finite, impact, intensity.point, quarter, stress
1 INTRODUCTION
O WING to high hardness, compressive and extremely

ly good corrosive resistances alumina is one of the
mostly used turbo materials. In last few decades few
heads have successfully withstood 107 cycles at maximum
walking stress of 17.2 KN, which is equivalent to a body
weight of 400Kg. The femoral heads didnt exhibit any
few of studies have been done on the dependency of sub-critical crack growth at the maximum walking load of
grain size on impact fatigue behavior of alumina.B. K 10KN, indicating the quasi-infinite performance life in-
.Sarkar and T.G.J Glinn (1969) studied the impact fatigue patient up to body weight of 250 Kg.
of an alumina ceramic and exhibit fatigue behaviors, hav-
ing a high stress plateau followed by progressively in- In recent past as new grade of alumina powder is availa-
creasing endurance with decrease in applied impact ble with very small particle size and new processing
energy. S Maity and B.K.Sarkar (1994) studied the impact routes are involved for new generation alumina products,
fatigue of a porcelain ceramic and showed a definite fati- which are coming in big way. Most of these new grades
gue behavior with increasing endurance in decreasing of alumina are of smaller grain size. It has been found
impact energy levels and cumulative residual stress is from the literatures that not too many studies on impact
suggested to explain the fatigue behaviors. S. Maity, D. fatigue behavior of alumina with fine grain size (particu-
Basu and B.K. Sarkar (1994) studied the fatigue behaviors larly sub-micron grain) have been done so far. So in the
of fine-grained alumina under repeated impact loading present work we have developed a machine, and using
and found that the fatigue resistance parameter is 17.12 that we have seen the effect of grain size on impact fati-
while the endurance limit is around 270Mpa which is gue behavior of alumina ceramics.
about 38% of the single impact strength of the material
and also found that fatigue cracks are trans-granular near
2 OBJECTIVES
the crack initiation region, the rest being inter-granular.
Manabu et al (2002) studied the material response to par- The objectives of the present study are as follows.
ticle impact during abrasive jet machining of alumina. A i) To develop an Impact Testing Machine for carry out the
relatively smooth face can be produced when silicon car- Experiment
bide (GC) abrasive is employed. ii) To see the Effect of Grain Size on Impact Fatigue beha-
The fatigue behaviour of fine-grained alumina hip joint vior
heads under normal walking load has also been reported iii) To study the dynamic stress intensity factor of Alumi-
by Basu et al (2005) and found that the alumina femoral na under impact loading,

Manoj Kumar Barai is currently pursuingan integrated Ph.D .program in
Jadavpur University, kolkata-32,West Bengal,India.
E-mail:[email protected]
JagabandhuShit is currently pursuingPh.D program in Jadavpur Univer-
3 EXPERIMENTAL METHODS
sity, kolkata-32,West Bengal,India. E-mail:[email protected] Sample preparation :Two grades of high purity commer-
cially available alumina powders with (i) average particle
IJSER 2011
130
ISSN 2229-5518
size of 180 nm, purity 99.99 % and (ii) average particle

size of > 0.4 m, purity 99.8 % have been used for this
study. The powders were used to obtain two different
grained dense compacts with average grain size of 0.4m Microstructure :
(Fig 1) and 4 m (Fig. 2). Alumina of different grain sizes
SEM photographs of the alumina samples of different
was prepared from different powders using different me-
grain sizes are shown below (Fig 3&,4).
thods as discussed below.
a. Samples were prepared in two different routes : Fig 3 Alumina samples sintered at 1275C following slip
(a) Isopressing followed by sintering and (b) slip casting with a grain size of 0.4 m
casting followed by pre-sintering and sintering
The process flow-sheets are given below.
The density of the sintered samples was measured using

water-immersion method. The sintered and ground sam-
ples were then polished for measuring various mechani-
cal properties like hardness, toughness etc. With the con-
ventionally prepared (iso-pressed and sintered ) samples,
hardness was found to be 16.0 GPa (average) and tough-
ness was 3.5-4 Mpam1/2 . The density was found to be
around 99% of the theoretical value.
In case of alumina made with slip casting process, density
was in excess of 98%, hardness was as high as 28 GPa and
toughness was comparable with that of the other grain
size.
Fig. 1. Process flow sheet for iso-pressed alumina samples

Fig 4 Alumina samples sintered at 1600C following iso-
pressing and pre-sintering with grain size 4 micron
With alumina with smaller grain size the grains are mo-
stly equiaxed where as for alumina with average grain
size, grains are little bit elongated in shape which enha-
nces grain bridging during crack propagation.A small set
up for measuring the impact fatigue behave-our has been
developed with a swinging pendulum of length of 54 cm
and a wight in the form a spherical ball of diameters
36mm. The impact load is given to the test specimen per-
Fig. 2. Process flow sheet for slip cast alumina
pendicular to its axis, which is rigidly fixed between two
IJSER 2011
131
ISSN 2229-5518
supports in a configuration resembling that of a Charpy experimental results it was observed that with the in-
test specimen. The bob (concentrated mass) is fixed at the crease of crack length the no of strike required for break-
end of the bar (swing arm), which is supported at almost ing the sample was decreasing for both the batch of sam-
frictionless hinge. The hinge is fixed inside the bearing, ples. But it was observed that even if the crack dimension
was more the no of strike required for breaking the sam-
which is mounted in the bracket plate. The angular move-
ple of grain size 4m was more than the sample of grain
ment is given to the set (drum & bob) by the cam mounted
size of 0.45m. This was probably due to the shape of the
on the low speed motor shaft. On the tensile surfaces of the
grains in higher grained samples which enhanced the
beam specimens made of two different grained alumina,
scope of grain bridging. In case of number of impacts re-
notches were created with different dimensions as men-
quired for breaking, there was variation in the pattern as
tioned below.
well.
Table 1: Number of strikes reqd to break vs. a/w for alu-

mina with grain size of 0.45 micron
Fig 5 Experimental setup for impact fatigue testing
Dynamic stress intensity factors were found out using

straight cracks on the tensile surfaces. Depths of the
straight cracks were varied. Depth of notch is given by a
and width of the specimen is given by w. a/w is the
non dimensional crack depth. This was done for both the Fig. 7. Variation of number of strike to break with non-
batch of alumina samples. It was not possible to produce dimensional crack length.
exactly similar notches but it was ensured that the depths
were comparable. From Fig. 7 it is evident that with alumina having lower
grain size, samples got broken with almost a single blow
in case of higher a/w (>0.025) while for the lowest a/w
(=0.01), there was a large scatter in the data. It shows that
with decrease in non-dimensional crack length (a/w),
influence of other intrinsic material features (e.g. presence
of other micro-structural discontinuities) got pronounced
resulting in high scatter from around 20 to 120. It was also
observed that a second degree polynomial equation nice-
ly fits the data with high (>95% fitting) percentage fitting.
Fig 6 (a) Three point bend specimen with straight crack,
(b) The range of crack depths in three point bend configu- Table 2. No of strikes to required to break vs a/w for
ration. alumina with grain size of 4 micron.
4. RESULTS & DISCUSSION
To perform the test we have taken samples of different

grain size such as 0.45m and 4m and it was observed
that the fatigue of ceramic samples of smaller grain size
was poorer than the higher grain size alumina. From the
IJSER 2011
132
ISSN 2229-5518
Fig. 8. Variation of number of strike to break with non-

dimensional crack length
Fig 9 Cracks propagating (sideways) from the existing

notch
For the samples with higher grain size, scatter was com-
paratively low, even the samples with a/w as high as 0.07
(even bigger than the largest a/w in the previous case)
could withstand around 80 impacts (average) before frac-
ture. With another set of specimens having a/w 0.097
(almost double the value of maximum a/w in the pre-
vious case), average number of impacts sustained before
fracture was around 30. Only one set having high a/w Fig. 10 Mixed mode of fracture in higher grained sample
(almost three times the maximum a/w of the previous
case) got broken with single impact.
From the SEM micrograph (Fig. 9), it is clear that the
cracks started propagating from the sides of the notches
and it followed a tortuous path. The crack apparently
followed a new plane after a reaching a particular point
or obstacle as observed on the fractured surface. In few
other cases, there was apparently a crack initiation site
(just at the tip of the notch) surrounded by semi-circular
arc-shaped zone. Fig. 10 shows that there was a mixed
IJSER 2011
133
ISSN 2229-5518
mode of crack propagation: both trans-granular and inter- [7]. John R., Stress intensity factor and compliance solutions for
granular. From the data available so far, it is evident that eccentrically loaded single crack geometry. Engineering Frac-
ture Mechanics (58) pp. 87-96, 1997
higher grained alumina samples showed superior impact [8]. John R. and Rigling B., Effect of height to width ratio on K
fatigue behaviour. The exact reason behind is not yet very and CMOD solutions for a single edge cracked geometry
clearly understood however a simple analysis of the two with clamped ends. Engineering Fracture Mechanics (60) No.
microstructures reveal that in case of higher grain size the pp 147-156, 1998
shape was a little bit elongated in contrast with nearly [9]. Kishimoto K., Aoki S. Sakata M., Dynamic stress intensity
factors using J-integral and finite element method. Engineer-
equiaxed shape observed with sub-micron grained alu-
ing Fracture Mechanics 13(2),387-394, 1980
mina. This caused grain bridging retarding partially the [10]. Maity S. Sarkar B.K, Impact fatigue of porcelain ceramic
propagation of cracks along grain boundary in case of International Journal of Fatigue 1995; 17(2), 107-109.
larger grained alumina. Furthermore with higher grain- [11]. Meggiolaro M. A., Miranda. A. C. O., Castro J.T.P., Martha L.
size chance of crack-branching could be more that could F., Stress intensity factors for branched crack growth. Engi-
reduce the energy available for the propagation of the neering Fracture Mechanics. 7 2 (2005) pp. 2647- 2671.
[12]. Nishioka T., Computational dynamic fracture mechanics,
main crack front. International journal of fracture 86 (1997) pp. 127-159
[13]. Rokach I. V., On the numerical evaluation of the anvil force
5. CONCLUSIONS: accurate dynamic stress intensity factor determination. Engi-
neering Fracture Mechanics 70(2003) 2059-2074
[14]. Rokach I. V., Estimation of the three-dimensional effects for
From the study it is evident that alumina with higher the impact fracture specimen. Arch Mech Engg 1996 ;43(2-
grain size showed superior impact fatigue behaviour in 3):241-252
[15]. Rokach I.V., Mixed numerical-analytical approach for dy-
comparison with that of sub-micron grained alumina.
namic one point bend test modeling., International journal of
Even with higher non-dimensional crack length (a/w), fracture 130,L193-L200 2004
alumina specimens with bigger average grain-size with-
stood higher number of impacts prior to fracture. The [16]. Sinclair G. B.,Messner T. W ., Meda G., Stress intensity fac-
superior impact fatigue behaviour was due to higher re- tors for deep cracks in bending. Engineering Fracture Me-
sistance to fatigue crack growth owing to grain bridging. chanics Vol. 55 No.1 pp. 19-24, 1996
[17]. Sinclair G. B., Meda G., Galik K., stress intensity factors for
ACKNOWLEDGMENTS side-by-side edge cracks under bending. Engineering Frac-
ture Mechanics Vol. 57 No.55 pp. 577-581,1997
[18]. Weisbrod G., Rittel D., A method for dynamic fracture
We acknowledge sincere help to the staff members of- toughness determination using short beams. International
Mechnical Engg Dept. & SBSE, Jadavpur University, Kol- Journal of Fracture (104) pp.89-103 2000
kata. One of the authors (M.K. Barai) sincerely ack-
nwledge constant support from the HOD, Director and
Principal of Future Institute of Engg and Management.
REFERENCES
[1]. Anderson, T.L., Fracture Mechanics: Fundamentals and Ap-

plication. 1995, CRC Boca Raton
[2]. Barsoum R.S. Triangular quarter- point elements as elastic
and perfectly plastic crack tip elements. In ternational Jour-
nal for Numerical methods in Engineering. (11),pp. 85-98
(1977
[3]. Bohme W, Kalthoff JF. The behavior of notched bend speci-
mens in impact testing. International Journal of Fracture
1982;20(4): R139-43
[4]. Chen Y.M. Numerical computation of dynamic stress intensi-
ty factors by a Lagrangian finite-difference method. Engi-
neering Fracture Mechanics 7(4), 653-660 (1975)
[5]. Enderlein, M., Ricoeur, A.,Kuna, M. Comparison of finite
element technique for 2D and 3D crack analysis under im-
pact loading. International Journal of Solids and Structures
40(13-14),3425-3437, 2004
[6]. Isida M., Effect of width and length on stress intensity factors
of internally cracked plates under various boundary condi-
tions. International Journal of fracture, 7, 301-316 (1971)
IJSER 2011
134
ISSN 2229-5518
Distributed Generation Planning Optimization

Using Multiobjective Evolutionary Algorithms
Mahmood Sheidaee, Mohsen Kalantar
Abstract In this paper, a method to determine the size - location of Distributed Generations (DGs) in distribution systems
based on multi objective performance index is provided considering load models. We will see that load models affect the
location and the optimized size of Distributed Generations in distributed systems significantly. The simulation studies are also
done based on a new multi objective evolutionary algorithm. The proposed method has a mechanism to keep the diversity to
overcome the premature convergence and the other problems. A hierarchical clustering algorithm is used to provide a
manageable and representative Pareto set for decision maker. In addition, fuzzy set theory is used to extract the best solution.
Comparing this method with the other methods shows the superiority of proposed method. Furthermore, this method can easily
satisfy other purposes with little development and extension.
istributed generation, Distribution systems, Load models, Strength Pareto Evolutionary Algorithm.
Index Terms D
1 INTRODUCTION
O ptimization was used to reconstruct electricity in-

dustry and looked for the best location for distri-
buted products. Development in technology and
2 LOAD MODELS AND IMPACT INDICES
client requirements to have cheap electric power and reli- To determine different load model effects on distributed
able one caused more motivation in distributed genera- generation planning, 37-bus distribution system will be
tion .Discussion about reliability and maintaining prevent studied (appendix 1)[7]. The effect of load models
the penetration of DG resources in the distribution sys- depends on voltage, means residential, industrial and
tems. commercial, will be studied in different planning
In [1] one approach was described based on genetic algo- scenarios. Load model defined as followed.
rithm for multistage planning of distribution systems op-
timizations. In this work, it's expressed as a mathematical Pi P0i Vi / Qi Q0i Vi (1)
model and algorithmic one and also tested with real sys- Where Pi and Qi are active and reactive power at bus i, P0i
tems. In [2] [5], it was studied on load models that are and Q0i are active and reactive power operating point in
usable for power flow and dynamic studies. This study bus I, Vi is voltage in bus i and and are active and
was done on load models depended on frequency or vol- reactive power exponents. In a constant power model
tage. During the recent years, studies on evolutionary conventionally used in power flow studied = = 0 is
algorithm have shown that these methods dont have the assumed. The values of the real and reactive exponents
difficulties of classical methods. In principle, multiple used in the present paper for industrial, residential and
Pareto optimal solutions can be found in one single run. commercial loads are given in Table 1 [7].
This paper has discussed on load model effects in location
and size planning and distributed generation optimiza- TABLE 1
tion. We can see that the load models affect on location EXPONENT VALUES
and size planning of DGs in distribution network. For the Load Type
purpose of studying on load models, its delivered loca- Constant 0 0
tion and size planning for single DG, its assumed that the Industrial 0.18 6.00
regarded DG has enough capacity. The proposed method Residential 0.92 4.04
is general and it can be used for case of multiple DG in Commercial 1.51 3.40
the network with increasing some variables.
This paper also suggested a new Strength Pareto Evolu-
During studying residential, its assumed that 38-bus
tionary Algorithm (SPEA) based approach for solving the
problem. The diversity-preserving mechanism embedded systems just has residential load. It's assumed that for
in the search algorithm makes it effective in exploring the industrial and commercial load, all loads are a kind of
problem space and capable of finding widely different industrial and commercial. In real situations, loads arent
solutions. A hierarchical clustering technique is imple- exactly residential, commercial and industrial, so the
mented to provide a representative and manageable mixture load class should be foreseen for distribution
Parto-optimal set. Also, a fuzzy-based mechanism has system. There are different ideas for studying DG effects
used the best solution for extraction. in distribution systems .One of this idea is different
IJSER 2011
135
ISSN 2229-5518
indices evaluation on the purpose of effect description on b) Search external Pareto set for the nondominated
distribution system because of DG during maximum individuals and emit all dominated individuals from the
power production. These indices are set.
1) Active and Reactive Power Loss Indices (ILP and ILQ): c) If the amount of the individuals externally stored in
P Q
ILP LDG 100 / ILQ LDG 100 (2)
the Pareto set exceeds a prespecified maximum size,
PL QL reduce the set by means of clustering.
Step 3) Fitness assignment: Calculate the amount of
Where PLDG and QLDG are total loss of active and reactive fitness values of individuals in both external Pareto set
power distribution system with DG, PL and QL are total and the population as follows.
loss of active and reactive power of total system without a) Assign appropriate each individual s strength
DG in the distribution network. amount in external set. The strength amount is
2) Voltage Profile Index (IVD): One of the advantage of proportional to the number of individuals covered by that
proper location and size of the DG is the improvement in individual.
voltage profile. b) The fitness of each individual in population is equal
to the sum of the strengths of all external Pareto solutions
which dominate that individual.
V1 Vi
IVD max in 2 100 (3) Step 4) Selection: combine the population and external
V1 set individuals. Choose two individuals randomly and

compare their fitness. Choose the best one and copy in a
3) MVA Capacity Index (IC): This informational index mating pool.
gives information in the field of system necessities for Step 5) Crossover and Mutation: do the crossover and
promoting transmission line. mutation according to new population production
probabilities.
S ij Step 6) Ending: check the ending criteria if all things are

IC max in2 (4) being done finish the work else substitute the old
CS ij population with the new one and go to step2.

In this paper, time searching will be stopped if the
generation counter exceeds its maximum number.
In some cases, the Pareto optimal set is extremely big or
3 PROPOSED APPROACH has extra solutions. An average linkage based hierarchical
Recently evolutionary algorithm showed that this clustering algorithm is used to reduce the Pareto set. We
algorithm can be effective for removing old method want to change P given set which its size exceeds the
problems [8]. The main element method of SPEA is maximum allowable size N to P* set with size of N.
1) External set: Its a set of Pareto optimal solutions. Algorithm is such as following [8].
These solutions were recorded externally and Step 1) Give primary amount to set C. each member of
continuously be updated. Finally recorded solutions P means a distinct cluster.
show Pareto optimal front. Step 2) if the number of clusters N, go to Step 5, else
2) Strength of a Pareto optimal solution: It is an assigned go to Step 3.
real value S[0,1) for each individual in the external set. Step 3) Calculate all the pairs of clusters distance. The
The strength of an individual is proportional to the distance dc of two clusters C1, C2 C is given as the
number of individuals covered by it. average distance between pairs of individuals across the
3) Fitness of population individuals: Fitness of each two clusters
individual in population is the sum of the strengths of all
external Pareto optimal solutions by which it is covered. 1
The strength of a Pareto optimal solution is at the same
dc
n1n 2 d i1, i2 (5)
i1c1 ,i 2c 2
time its fitness.
Algorithm is in the following steps [8].
Where n1 and n2 are clusters individuals of C1 and C2.
Step 1) primary amounts: produce population and make
Function d shows Euclidian distance between i1 and i2.
empty external Pareto optimal set.
Step 4) Determine two clusters that have minimum dc
Step 2) updating external set: External Pareto optimal set
distance. Combine these clusters into a larger one. Go to
is updated as following:
Step 2.
a) Search population for the nondominated individuals
Step 5) find centroid for each cluster and choose the
and copy them in the external pareto set.
nearest individual to the centroid as agent and emit other
individuals from the cluster.
IJSER 2011
136
ISSN 2229-5518
Step 6) Compute the reduced nondominated set P* by Where is a binary random number, r is a random
uniting the representatives of the clusters. number r[0,1], gmax is maximum number of generations
As soon as having the Pareto optimal set of and is a positive constant that is desirable. = 5 is
nondominated solution, the proposed approach presents selected. This operator gives a value xi[ai,bi] such that
one solution as the best compromise solution. Each the probability of returning a value close to xi increases as
objective function of the i-th solution is represented by a the algorithm advances. This makes uniform search in the
membership function i defined as initial stages where t is small and for later stages is so
local.
1
F max F Fi Fimin
i i 5 MULTIOBJECTIVE BASED FORMULATION
i F min Fi Fimax (6)
max min i Multiobjective index for evaluating distribution systems
F
i F

i Fi Fimax operation on purpose of DG location and size planning
0
with load models, considers all previously mentioned
For each nondominated solution, the normalized indices by strategically giving a weight. The
membership function k is multiobjective index operation on basis of SPEA
algorithm is according to (10).
N obj
IMO 1 .ILP 2 .ILQ 3 .IC 4 .IVD (10)
ik
i 1
k (7) These weights are because of giving the corresponding
M Nobj
importance to each impact indices. Table 2 identifies used
ik
amount for the weights with regarding normal operation
k 1 i 1
analysis [7].
where M is the number of nondominated solutions. The TABLE 2
best solution is the one that has more k. INDICES WEIGHTS
. Indices p
ILP 0.40
4 IMPLEMENTATION OF THE PROPOSED APPROACH ILQ 0.20
IC 0.25
Because of Binary representation problems when search
IVD 0.15
space has wide dimension, the proposed approach has
been implemented using Real Coded Genetic Algorithm
(RCGA). Decision variable xi has real amount within limit Multiobjective function (10) can be minimized with
of ai and bi (xi[ai,bi]). The RCGA mutation and crossover regarding to various operational constraints to satisfy the
operators RCGA is like this. electrical requirements for distribution network. These
Crossover: A blend crossover operator (BLX-) has been limitations are:
employed in this paper. This operator will choose one 1) Power Conservation Limits: The algebraic sum of all
number randomly from the interval [xi (yi - xi) , yi + input and output powers, such as distribution network
(yi - xi)], where xi and yi are the ith parameter values of total losses and power generated from DG, which should
the parent solutions and xi < yi. Because of ensure the be equal with zero. (NOL = no of lines)
balance between exploitation and exploration from search
space, = 0.5 is chosen. n NOL
Mutation: Nonuniform mutation was used here. In this Pss i, V PD i, V Ploss V PDGi (11)
operator, new amount xi of parameter xi produced after i2 n 1
mutation in t time.
2) Distribution Line Capacity Limits: Transmission
capability in each line should be equal with thermal
x t , bi x i if , 0
xi i (8) capacity.
x i t , bi x i if , 1
S i, j S i, j (12)
max

1 t
g max
t , y y 1 r (9) 3) Voltage Drop Limits: voltage drop should base on
voltage regulation that DISCO gives.

IJSER 2011
137
ISSN 2229-5518
Fig. 1. Impact indices and IMO with DG size-location pair for

V1 V j V max (13) constant load
If voltage and MVA limits in system buses for a par-

ticular size and location, accept that pair for next genera-
tion, else this size and location will be ignored and re-
jected. Size and location should be had minimum IMO.
6 SIMULATION RESULTS
The multiobjective index based analysis is carried out on
37-bus test systems as given in the Appendix [7]. A DG size
is considered in a practical range (00.63 p.u.). It's assumed
that DG is operated at unity p.f.. This assumption has two
reasons:
1) Usually when the DG has unity power factor, has
maximum profit because the cost of active power is higher. Fig. 2. Impact indices and IMO with DG size-location pair for
Use at unity power factor cause to have maximum capaci- industrial load
ty. The solution obtained using constant power load
2) Used models in this paper are simple and more atten- models may not be feasible for industrial load. A similar
tion is for voltage changes dependence of load models. and significant effect of load models can be easily be
The using method hasnt been limited by DG models observed from the Figs. (3) (5) for residential
and it's general. First bus was chose as feeder of electric commercial and mixed load models. The differences in
power from network and the rest buses are regarded as DG values of DG size, IMO and its components are
location. On all optimization runs, the population size and significant, showing that the load models effects are
maximum number of generations were selected as 200 and important for suitable planning of size and location. Table
500, respectively. The Pareto optimal set maximum size 3 summarizes the optimal DG size-location pairs, IMO
includes 20 solutions. The crossover and mutation proba- along with its components for each kind of load. From
bilities were selected as 0.9 and 0.01, respectively. For 37- Table 3, the optimal size-location for constant load model
bus system, variation of impact indices and IMO have been (0.6299 p.u. bus 14) is different with industrial load
shown with DG size and location in figure 3-7 for constant, model (0.63 p.u. bus 14) residential load model (0.4672
industrial, residential, commercial and mixed load models. p.u. bus 14) commercial load model (0.4419 p.u. bus
The value of IVD for all load models is near zero. It shows
14) and mixed load (0.5113 p.u. bus 32). Similarly IMO
that voltage profile improves with present DG.
and other effective indices for optimal DG location-size
We can see from Figs. (1) (5) that the indices ILP, ILQ,
are different.
IC and IMO achieve values greater than zero and smaller
than one, indicating the positive impact of DG placement
in the system. Fig. 1 shows that values of IC, ILP and ILQ
for buses 2-4 as IC<ILP<ILQ and for buses 6-8 like
ILQ<ILP<IC. Figure 2 shows the value of optimum DG
size, IMQ and its components for all buses for industrial
load model. So load models affect on solutions.
Fig. 3. Impact indices and IMO with DG size-location pair for

residential load
IJSER 2011
138
ISSN 2229-5518
TABLE 4
COMPARISON OF SYSTEM POWER LOSSES AT
OPTIMAL LOCATION OF DG WITH LOAD MODELS
Load Optimal PLDG PL 0.01 QLDG QL 0.01
model location 0.01 p.u. p.u. 0.01 p.u. p.u.
Constant 14 0.1499 0.2002 0.0991 0.1335
Industrial 14 0.1464 0.1671 0.0968 0.1112
Residential 14 0.1358 0.1604 0.0896 0.1066
Commertial 14 0.1166 0.1548 0.0767 0.1028
Mixture 32 0.1142 0.1588 0.0766 0.1056
Fig. 4. Impact indices and IMO with DG size-location pair for 6.3 Conclusion
commercial load
The general analysis includes load models is proposed for
locationsize of distributed generation planning in multiob-
jective optimization in distribution systems. The multiobjec-
tive criteria depends on system operation indices is used in
this work. It was seen that while regarding load models,
there will be changed in DG location and size. The overall
value of multiobjective index (IMO) changed during charge
model changing.
Also in this paper, we suggested a new method based on
Pareto evolutionary algorithm and used for DGs location
size planning problem. This problem formulized as a mul-
tiobjective optimization problem, A diversity preserving
mechanism for finding widely different Pareto optimal solu-
Fig. 5. Impact indices and IMO with DG size-location pair for mixture tions was used. A hierarchical clustering technique is im-
load plemented to provide a representative and manageable Pa-
The probable DG location-sizes may be little (because reto optimal set without destroying the characteristics of the
of constraints), but the number of candidate solution are trade-off front and a fuzzy based mechanism is used for
fairly large to suggest the application of SPEA. The finding the best compromise solution. The result shows that
differences in values of DG size, IMO and its components the suggestive method for multiobjective optimization prob-
are significant for load models, showing that the load lem is useful, because multiple Pareto optimal solutions are
models effects are important for suitable planning of size found during simulation. Since the proposed approach does
and location. The values of QLDG and PLDG related to not impose any limitation on the number of objectives, its
optimal size location for any kind of load model have extension to include more objectives is a straightforward
process.
been shown in table 4, although the values of QLDG and
PLDG for nonconstant load models (industrial residential APPENDIX
commercial and mixture) arent different but their
Fig. 6 shows the 37-bus test system.
difference is significant when compared to constant load
model.
TABLE 3
IMPACT INDICES COMPARISON FOR PENETRATION
OF DG WITH LOAD MODELS
Indices Constant Industrial Residential Commercial Mixture
ILP 0.7078 0.6517 0.7459 0.7756 0.7526
ILQ 0.7035 0.6449 0.7383 0.7685 0.7551
IC 0.9913 0.9671 0.9570 0.9476 0.9478
IVD 0.0687 0.0634 0.0661 0.0653 0.0696
IMO 0.6823 0.6409 0.6952 0.7106 0.6994
Location 14 14 14 14 32
Size 0.6299 0.63 0.4672 0.4419 0.5113
IJSER 2011
139
ISSN 2229-5518
REFERENCES
[1]. V. Miranda, J. V. Ranito, and L. M. Proenca, Genetic algorithms
in optimal multistage distribution network planning, IEEE
Trans. Power Syst., vol. 9, no. 4, Nov. 1994, pp. 19271933.
[2]. C. Concordia and S. Ihara, Load representation in power sys-
tems stability studies, IEEE Trans. Power App. Syst., vol. PAS-
101, no. 4, Apr. 1982, pp. 969977.
[3]. IEEE Task Force on Load Representation for Dynamic Perfor-
mance, Bibliography on load models for power flow and dy-
namic performance simulation, IEEE Trans. Power Syst., vol.
10, no. 1, Feb. 1995, pp. 523538.
mance, Load representation for dynamic performance analy-
sis, IEEE Trans. Power Syst., vol. 8, no. 2, May 1993, pp. 472
482.
Fig. 6. 37-bus test system
mance, Standard load models for power flow and dynamic

performance simulation, IEEE Trans. Power System, vol. 10,
no. 3, Aug. 1995, pp. 13021313.
[6]. C. A. C. Coello, A comprehensive survey of evolutionary
based multiobjective optimization techniques, Knowledge and
Information Systems, vol. 1, no. 3, 1999, pp. 269308.
[7]. E D. Singh, D. Singh, K. S. Verma, Multiobjective Optimization
for DG Planning With Load Models, IEEE Trans. Power Syst.,
vol. 24, no. 1, Feb. 2009, pp. 427-436.
[8]. M. A. Abido, Environmental/Economic Power Dispatch Using
Multiobjective Evolutionary Algorithms, IEEE Trans. Power
Syst., vol. 18, no. 4, Nov.2003, pp. 15291537.
[9]. C. M. Huang, H. T. Yang, and C. L. Huang, Bi-Objective power
dispatch using fuzzy satisfaction-maximizing decision ap-
proach, IEEE Trans. Power Syst., vol. 12, Nov. 1997, pp. 1715
1721.
[10]. D. B. Das and C. Patvardhan, New multi-objective stochastic
search technique for economic load dispatch, Proc. Inst. Elect.
Eng.-Gen. Transm. Dist., vol. 145, no. 6, 1998, pp. 747752.
[11]. M. E. Baran and I. M. El-Markabi, A multiagent-based dis-
patching scheme for distributed generators for voltage support
on distribution feeders, IEEE Trans. Power Syst., vol. 22, no. 1,
Feb. 2007, pp. 5259.
[12]. E. G. Carrano, L. A. E. Soares, R. H. C. Takahashi, R. R. Saldan-
na, and O. M. Neto, Electric distribution network multiobjec-
tive design using a problem-specific genetic algorithm, IEEE
Trans. Power Del., vol. 21, no. 2, Apr. 2006, pp. 9951005.
IJSER 2011
140

ISSN 2229-5518
Marketing Of Asian Countries as

Tourist Destination- Comparative Study
of India and Malaysia
Dr Reshma Nasreen, Nguyen Toan Thang
Abstract - Tourism has gradually grown over the years as a full fledged industry. Many countries are gaining from this
welcome change. The contributions of this sector to the countrys coffers are sizable for some countries, while some
countries have a long way to go. This research paper attempts to study the reasons of lack of optimal contribution of this
sector in India and also forays into strategies that can be adopted to capitalize on the patterns prevalent in tourist
behavior. A country like India with a commendable historical significance and size has not been able to garner as much of
tourist attention because of certain factors. India has a lot of offerings to whip the appetite of an avid tourist, but the
varieties have either not been promoted, or if promoted lack of associated services have not led to desired synergies.
After identifying the gaps between the two countries (India & Malaysia), the paper puts forth the tourists patterns of
behavior through the data collected. The questionnaire has been administered to tourists in New Delhi and Agra (cities in
India). Malaysia on the other hand has had a steady stream of tourists trickling down and benefiting its economy.
Key Words- ASEAN, Eco-tourism, Heritage Sites, MICE, Ministry Of Tourism (India & Malaysia). World Travel.
INTRODUCTION the World Tourism Organization (WTO)

for 1996 show that tourist arrivals and
Tourism plays an important role in the
economies of a number of ASEAN
receipts accounted for a 15.2% and
19.4% share of the world's total
as well as other Asian countries. Besides respectively, a significant increase from
contributing to the national income, 1985 as shown in Tables I and 2 (WTO
promotion of intra-regional tourism has 1997a).
other beneficial spillovers for trade and
people-to-people contacts. The shared In his paper, Asia Pacific
history and culture dating back to several Tourism Industry: Current Trends and
centuries provides a base for tourism Future Outlook, Mr Amrik Singh points
exchanges. A large number of tourists out, that the rapid growth of the tourism
from Southeast Asia come to centre of industry has been attributed to a number
Buddhist pilgrimage in India every year. of factors including among others, strong
With the rise of Indian middle class with economic growth, increase in disposable
higher purchasing power, India has also income and leisure time, easing of travel
emerged as a big market for ASEAN restrictions, successful tourist
countries as a source of tourists. promotion, and a recognition by the host
According to Amrik Singh, governments that tourism is a powerful
instructor at Department of Parks, engine of growth and a generator of
Recreation and Tourism, University of foreign exchange earnings.
Utah, The Asia and Pacific region will
be the focus of the worldwide tourism
industry in the new millennium. Over
the last decade, tourist arrivals and
receipts rose faster than any other region
in the world, almost twice the rates of
industrialized countries. Statistics from
IJSER 2011
141

ISSN 2229-5518
It is expected that expenditure by

Purpose of Visit % of Tourists international tourists in Malaysia
Leisure, Recreation and 45.45 will increase at a CAGR of
Holiday 6.63% during the forecasted
Visiting Friends and 12.01 period.
Relatives Increasing disposable income in
Business and 23.00 Malaysia will open the
Professional opportunities for both outbound
Health and Treatment 2.20 and domestic tourism. It is
Religion and Pilgrimage 12.24 expected that per head disposable
Others 5.10 income in the country will
Total 100 increase at a CAGR of 5.06%
Fig 1: Purpose Of Visit to India during 2008-2012increase at a
(Source: Ministry Of tourism, CAGR of 5.06% during 2008-
Government of India) 2012.
It is expected that MICE
(Meetings, Incentives,
Conventions & Exhibitions)
As per reports in Opportunities in industry will be one of the major
Malaysian Tourism Industry (2007- contributors to the Malaysian
2009) tourism industry.
http://www.marketresearch.com/product/
display.asp?productid=1806312, the key Ministry of Tourism, Government of
findings include the following. India in its Annual Report, has
recognized the need for further
Singapore, Thailand and promoting India through Incredible
Indonesia are important sources India campaign. The ministry aims
of visitors for Malaysia. Beyond to capitalize on the rich and varied
ASEAN, tourist arrivals from history of India as well as it being
China and India will remain an the centre of origin of two important
important influence throughout religions, Hinduism and Buddhism.
the forecast period (2008-2012) Apart from the fact that India has a
as the majority of Chinese tend to rich history, at present it is also
weigh their spending towards becoming a preferred location for
consumer purchases as opposed medical tourism as well as a
to luxury hotel accommodation. recognized centre for higher
The promotion of Education education. The vast coastline of
Tourism will continue to be India, along with the charms of
expanded to expedite the Northeast India beckons tourists who
development of Malaysia as a are nature lovers. The purpose of
preferred destination for visit of tourists coming to India has
international students. The been summarized in the Table 2:
projected foreign exchange
earnings from this potential Although around 50 percent of the
source of growth are estimated at tourists coming to India go for
RM 900 Million by 2010. leisure and recreation, many
IJSER 2011
142

ISSN 2229-5518
attractive destinations can be Table 2: Comparison of Attractions

promoted apart from the beaches of and Infrastructural support between
Goa (a state in Western India) and India and Malaysia
Andaman and Nicobar Islands (chain
of Islands in the Bay of Bengal,
towards the east of India). In the Yea Malays Malays India India
r ia ia Arrival Receip
following table some of the factors Arrivals Receipt s ( ts
attracting tourists as well as (million) s (USD) million (USD)
facilitating them have been billion ) billion
compared between the two countries. 200 5.363 2.283 3.016
As can be clearly seen from the Fig 1 12.7
2, opportunity for attracting tourists 200 7.502 2.073 3.012
2 13.2
is comparatively more in India than 200 2.726 5.145
7.967
Malaysia. Throughout the length and 3 10.5
breadth of Malaysia, not much 200 6.572 3.457 5.588
cultural diversity can be witnessed. 4 15.7
India on the other hand is known for 200 9.176 3.919 6.624
5 16.4
its diversities and different cultures.
200 9.889 4.447 7.805
India has got more than double the 6 17.4
number of international airports, nine 200 11.222 4.977 8.872
times more the number of heritage 7 20.9
sites, three times more the number of 200 14.322 5.287 10.146
beaches, four times more railway 8 22.1
200 15.872 5.113 `10.99
connectivity and ten times the area of 23.6
9 2
Malaysia. But still the number of
tourists coming to India is Table 3: Comparative chart of
substantially less than the numbers arrivals and receipts of Tourists in India
coming to Malaysia as is shown in and Malaysia
Table 3.
(Source: Ministry of tourism,
Malaysia; Ministry of Tourism India)
Parameters Malaysia India
No Of 5 11
International
Airports
Cultural 3 28
Heritage Sites
No of 7 21 METHODOLOGY OF DATA AND
Beaches ANALYSIS
Railway 1,.699 km 64,015 km
connectivity Scope of the Study
Uniqueness Eco- Historical Tourism is one industry which
tourism and has been actively promoted in India as
religious well as Malaysia. India went on to
Area 328,600 3,287,263 promote itself as a tourist destination
sq.km sq. km through the Incredible India campaign,
which was launched at different film
festivals around the globe as well as at
IJSER 2011
143

ISSN 2229-5518
national level through advertisements - To evaluate the performance of

promoting the different states of India. existing tourist facilities in the country.
Malaysia on the other hand has been - To assess the number of
promoting itself as a tourist destination outbound tourists to various countries.
through the Malaysia-Truly Asia - To undertake market
campaign. segmentation analysis.
Both these countries are - To estimate the average
developing countries and Asian duration of stay of foreign tourists in
countries. The study aims to study the India.
difference in the purpose of visit or - To obtain demographic,
similarities in the tourists decision economic and social profiles of foreign
making process when he/she opts for this tourists visiting India and the
as a vacation spot. The study would also motivational factors responsible for
be of help to researchers and decision attracting them to India.
makers as it is both informative and
instructive.
Objectives of the study Research Design

The survey was conducted with
the following objectives: Descriptive: Descriptive statistics has
- To assess demand profiles of been used to describe the basic features
the foreign tourists visiting the county, of the data in this study. Simple
this may include demand for tourist summaries about the sample and the
places, tourism goods and services measures have been provided. With
including transport and accommodation, simple graphics analysis, the quantitative
travel agencies, cultural services, data has been analyzed.
recreation and other entertainment
services. Sources of data
- To assess the detailed Secondary: Annual report of Malaysia
expenditure pattern of foreign tourists Tourism Ministry 2006 2008
visiting India, and estimates of total Annual report of India
expenditure at all India level on specific Tourism Ministry 2006 2008
tourism products such as Annual report of World
accommodation, food & beverage Travel and Tourism Council
services, passenger transport, tour Incredible India report
operators and tourist guide services, 2006 - Tourism Ministry of India
transport equipment on rental, cultural
services, recreation and other Primary: Questionnaire close ended
entertainment services and other tourism and multiple choice questions.
related products and services.
IJSER 2011
144

ISSN 2229-5518
America (USA, Canada, Haiti, Trinidad &

Research Instrument Tobago) 7%
Questionnaire close ended and multiple
choice questions were asked European (England, Spain, German, Ukraine,
Questionnaire attached as annexure
Lithuania, Bulgaria, Romania) 12%
Northeast Asia (China, Japan, South Korea,
Sampling Method: Restricted random
Mongolia) 21%
sampling
Middle East (Iran, Iraq) 16%
SAMPLE SIZE South Asia ( Sri Lanka, Afghanistan, Nepal) 17%

A sample of 100 tourists was surveyed. Africa (Nigeria, Yemen, Uganda, Chad) 9%
The sample foreign tourists were
contacted for collection of detailed ASEAN (Thailand, Indonesia, Vietnam,
information in a structured schedule at 2 Malaysia, Singapore) 18%
exit points Delhi and Agra.
Total 100%
Limitations of the study
Asia countries 17% and Middle East
Language Barrier: Some of the tourists,
they come from South Asia countries 16%.
likes Sri Lanka, Bangladesh, Nepal can
speak Hindi very well but they do not
know English.
Time Constraint: Researcher had only
few weeks to do this survey
questionnaire, one week in New Delhi
and one week in Agra. The availability of
foreign tourists for interview was
constrained by the time at their disposal
Sample Constraint.
SURVEY RESULT
I. DEMOGRAPHICS
Nationality
An estimated 100 foreign tourists
come to India from various country
covered in the survey during March
2010. The major countries accounted for
39% of foreign tourist are East Asia, out
Gender
of America 7%, European 12%, South
IJSER 2011
145

ISSN 2229-5518
Nearly 62% of foreign tourists

who visited India were males. The ratio
varied from country to country.
Male 62
Female 38
Total 100
Age
The tourists were classified into
seven-age groups viz., upto seventeen, Education
eighteen to twenty four, twenty five to The tourists were also classified on the
twenty nine, thirty to thirty four, thirty basis of educational levels. The survey
five to thirty nine, forty to forty four and reveals that nearly 67% of the foreign
forty five to forty nine. Nearly 60% of nationals visiting India were graduates
the tourists belonged to the age-group and postgraduates at higher education or
eighteen to thirty, the next highest group university; only 3% tourist at lower
was thirty to thirty five (14%). vocational education.
17 years or younger 1% Lower (incl. lower vocational education) 3%

18-24 years 30% Secondary education (general) 10%
25-29 years 31% Secondary vocational education 20%
30-34 years 14% Higher education or university 67%
35-39 years 13% Total 100%
40-44 years 9%
45-49 years 2%
More than 50 years 0%
Total 100%
IJSER 2011
146

ISSN 2229-5518
Total 100%
III. MOTIVATIONAL FACTORS

The survey shows that 67%
tourists visited for leisure, recreation &
holiday, 15% for business, study purpose
and 11% for religion and pilgrimage, 2%
visited to meet friends and relatives in
India, remaining 5% visited India for
other purposes.
II. SOURCE OF INFORMATION
Family/friends are major (30%)
source of information, followed by What was the main purpose of your
Internet (22.4%), travel agents/tour holiday to India?
operators (16%), travel Round trip 11%
magazines/films/T.V. etc. (4%) and Festivals 13%
other (13.6%). Eco-tourism 6%

Nature holiday 12%
Where did you get the Beach holiday 5%
information about this destination Cultural holiday 20%
Travel Agency 16% Spiritual holiday 11%
Family/friends 32% Family visit 2%
Spa / wellness
Internet 22.40%
Active holiday
Papers/Magazines 4%
Honeymoon
Travel Brochures 1.60%
Study / placement /
TV/Radio 4%
work 15%
Tourism board 0.80%
Other: 5%
Travel guides 5.60%
Total 100%
Other 13.60%
IJSER 2011
147

ISSN 2229-5518
V. EXPENDITURE PATTERN
The analysis of tourist
IV. TRAVEL PATTERN expenditure shows that 23% tourists
The analysis of travel pattern spent around 1000 USD, 19% tourists
shows that 30% of tourists traveled spent around 750 USD, 14% spent
alone, 20% traveled with two persons, around 500 USD and only 8% tourist
22% traveled in a group of 3 persons, spent above 2250 USD.
16% in a group of four persons, 12 % in What were the travel and
a group of five persons and more. lodging expenses of this trip to
India per person?
1 person 30% Around USD 500 16%
2 persons 20% Around USD 750 19%
3 persons 22% Around USD1.000 23%
4 persons 16% Around USD 1.250 14%
5 persons or more 12% Around USD 1.500 8%
Total 100% Around USD 1.750 7%
Around USD 2.000 5%
USD 2.250 or more 8%
Total 100%
IJSER 2011
148

ISSN 2229-5518
From data survey we find that 37% of

VI. TRANSPOST USED MOST IN
tourists spent 20 USD per day in India
INDIA
excluding travel and lodging expenses
The major mode of travel within
and 18% of tourist spent 30 USD per
India used by foreign tourists shows that
day, 15% of tourists spent 40 USD, 18%
almost one-third used Rail, remaining
of tourists spent around 50 USD and
one-third used Road and 10% used air
12% of tourists spent more than 60 USD.
transportation.
What means of transport did
Spend per person per day in India,
you use most in India?
excluding travel
and lodging expenses Car (for rent) 4%
Around USD 20 37% Airplane 8%
Around USD 30 18% Train 31%
Around USD 40 15% Rickshaw 21%
Around USD 50 18% Taxi 10%
USD 60 or more 12% Bus (public transport) 9%
Total 100% Motorcycle (rent) 1%
IJSER 2011
149

ISSN 2229-5518
Bus (excursion) 9% 22 - 30 days 22%

Bike 3% 1 - 2 months 19%
Other: 4% 2 months or more 10%
Total 100%
Total 100%
VIII. ACCOMMODATION
VII. NUMBER OF DAYS STAYED
There are 60% tourists stayed in
There are 68% tourists have
middleclass hotel; 3% spending in
more than one week to four weeks
luxury hotel (4 and 5 stars). 12% tourists
stayed in India, only 3% of tourists
come to their friends, relatives or family.
stayed one week or less. Around 30%
tourists stayed more than one month to
At what kind of accommodation did you
two months in India.
stay in India?
Luxury hotel (4 and 5 stars) 3%
How long do you have holiday in
Middleclass hotel (3 stars and
India?
less) 37%
7 days or less 3%
Guesthouse 23%
8 - 14 days 18%
Apartment / bungalow 4%
15 - 21 days 28%
Private home / villa 3%
IJSER 2011 Friends / relatives / family 12%
Other 18%
Total 100%
150

ISSN 2229-5518
How many times did you visit

India before?
Not yet 33%

Once 27%
Twice 19%
Three times 11%
Four times 8%
Five times or more 2%
Total 100%
IX. REPEAT VISITS
Yes 53%
No 27%
Do Not Know 20%
Total 100%
While 33% of the tourists visited

India for the first time, for 27% it was
the second visit, 19% third visit and 2%
of them had visited India more than five
times on earlier occasions.
X. VALUATION OF STAY IN INDIA

The tourists were asked to evaluate their overall
satisfaction level as Most satisfying, Satisfying, Average
evaluated their overall visit to India as Most Satisfying
Average and only 14% rated it as Disappointing and most d
IJSER 2011
151

ISSN 2229-5518
What is your valuation of your India, remaining 11% visited India for
stay in India? other purposes.
Most satisfying 15% What would be the main purpose of
Satisfying 28% your next visit to India?
Average 31% Round trip 12%
Dissatisfying 12% Festivals 8%
Most dissatisfying 2% Eco-tourism 4%
DK/NA 12% Nature holiday 12%
Total 100% Beach holiday 7%
Cultural holiday 13%
Spiritual holiday 12%
Family visit 7%
Spa / wellness
Active holiday 4%
Honeymoon
Diving holiday
Study / placement / work 10%
Other: 11%
Total 100%
XI. MAIN PURPOSE OF NEXT

VISIT TO INDIA
The survey shows that 56%
tourists visited for leisure, recreation &
holiday, 10% for business, study purpose
and 12% for religion and pilgrimage, 7%
visited to meet friends and relatives in
IJSER 2011
152

ISSN 2229-5518
XII. FAVOURITE HOLIDAY XIII. WOMAN TRAVEL PATTERN

DESTINATION COUNTRIES IN In 38% of woman tourists to
ASIA IN THE NEXT FEW YEARS India, only 18.42% one person comes to
In total 181 ideas show out, Japan got 38 India, more than 80% of them come to
ideas, Korea South with 26, China with India with group of two, three, four, five
25, Hong Kong had 20 and Malaysia got persons.
17 ideas - East Asian countries are We find that tourists feel not
favorite holiday destinations of tourists. safe, harassment. The inflow of the
Favorite holiday destination countries in foreign tourists could be ensured only
Asia in the next few years when they are provided safe and secure
Japan 38 environment.
China 25
Thailand 14
Female travel pattern
Singapore 12
1 person 18.42%
Hong Kong 20
2 persons 13.15%
Malaysia 17
3 persons 21.07%
Indonesia 11
4 persons 23.68%
Korea South 26
5 persons 23.68%
Saudi Arabia 3
Total 100.00%
UAE 2
Israel 4
Vietnam 6
Srilanka 3
Total 181
IJSER 2011
153

ISSN 2229-5518
XIV. MAIN PURPOSE OF NEXT Short Term Measures

Main purpose of next visit to India Finding: One
- tourists under 30 years old i. A vast majority of respondents (63%)
Round Trip 19.35% had reported over charging, by taxi
Festivals 11.29% drivers, lack of manners, ill-informed
Eco-Tourism 6.45% tourist guides and tour operators,
Nature holiday 17.75% misbehavior with ladies, cheating, etc in
Beach holiday 9.67% many tourist destinations like Agra,
Cultural holiday 16.12% Delhietc.
Spiritual holiday 4.85% Suggestions
Study/ placement/work 9.67% i. The police should be given adequate

powers and additional staff to enforce
Other 4.85%
discipline and change the mind set of
Total 100.00%
different players in the tourism sector.
State governments may examine the
VISIT TO INDIA TOURISTS
feasibility of constituting a special
UNDER 30 YEARS OLD
tourist police force to position at
We find the trend of more than
different tourist centers/ tourists spots,
80% tourists under 30 years old would
monuments.
like to come to India for leisure,
ii. A code of conduct should be framed
recreation & holiday; 4.85% for spiritual
and publicized for tourist guides, taxi
holiday purpose; around 10% come to
drivers, tour operators, travel agents etc.
India with study, business purpose.
Concerned associations should be
partnered with this proposed activity.
Finding: Two
i. About 90% tourists remarked that
there were inordinate delays at
immigration counters. These delays were
attributable to lack of space and
inadequate computerization, coupled
IJSER 2011
154

ISSN 2229-5518
with heavy rush during morning hours i) A comprehensive study should be

(midnight to 3-4 A.M.) commissioned by the Ministry of
Suggestion Transport in collaboration with
i. The authorities concerned at Department of Tourism to ascertain the
international airports must ensure status of approach roads and prepare a
availability of more space so as to enable time bound action plan to rectify the
the Ministry of Home Affairs (MOHA) defects to make them motorable and
to locate additional counters and deploy travel friendly in major tourist
more immigration officers. destinations.
ii. The paper work involved in Finding: Four
immigration should be reduced in line i) The affluent countries of Western
with international practice i.e; all Europe, North America, Japan and
outgoing passengers as well as incoming Australia are very far from India.
nationals (Indian citizens) should not Several foreign nationals (about 30%)
have to fill in an elaborate form, and cited that high cost of international travel
incoming foreigners should also be stood in the way of attracting a larger
required to fill in only simple forms that number of tourists from these countries.
are amenable to easy and fast processing The only way to reduce per capita cost
for generation of essential input for of international travel is through group
concerned authorities like Home Affairs, tours and charter flights.
DOT, Civil Aviation etc. Suggestion
i) Recent steps taken by the government
of India by 1) Abolition of the inland air
Medium / Long term Measures travel tax, 2) scrapping the basic fare and
Finding: Three foreign travel tax of Rs. 500, 3) The
i) Sizeable percentage of tourists (55 %) reduction in excise duty on aviation
opined that most of the approach roads turbine fuel to 8% from 16% etc will go
in tourist destinations were in bad a long way in bringing down the cost of
conditions causing great inconvenience travel. Similarly the state governments
and delays. should relook at exorbitant luxury taxes,
Suggestion sales tax etc. This will substantially
IJSER 2011
155

ISSN 2229-5518
reduce package tour cost thus increasing Suggestion

the flow of tourists. Overseas and i) Uniform fares for foreign and Indian
domestic tour operators should be nationals should be charged so that wide
encouraged to organize package tours spread feeling of discrimination is
with a focus on specific tourism products reduced.
like medical tourism, spiritual tourism, Finding : Seven
eco tourism etc. i) India has not succeeded in attracting
ii) Efforts must be made to draw ethnic family groups to visit India together in
groups other than Indians from large numbers as is evident from the
Southeast Asian countries to places in very low percentage of children,
India with Buddhist relics and scenic and housewives and females amongst the
cultural attractions. Steps initiated tourists (38% woman). There is a great
recently should be intensified and spread need for attracting tourists to visit India
throughout the region. Opening of new for the pure pleasure she offers.
air connectivity under the Open Skies Historical sites, palaces and architectural
Policy and the SAARC agreements monuments, hills, beaches and forests,
should facilitate such flights. wildlife, religious and folk festivals,
Finding: Five music and dances film and theatre,
i) The awareness of unique tourism handicrafts of India must be all great
products like Health tourism facilities attractions to foreigners.
are very poor. Suggestion Suggestion
i) Developing suitable persons/ agencies i) Group tourism should be encouraged
to bring out directories / brochures on by devising suitable strategies and
unique tourism products, like Health promotional measures to attract younger
tourism, and distribute to Indian generations, housewives etc. About 58%
missions abroad, tour operators, travel of the tourists were repeat visitors.
agents etc. in the target source markets. Finding: Eight
Finding: Six i) More than 30 % of the tourists visit
Entry fees to monuments and heritage India on their own motivation or at the
sites are higher for foreigners than influence of others who had visited India
collected from the Indian Nationals earlier.
IJSER 2011
156

ISSN 2229-5518
Suggestion tourism product and lay more emphasis

i) It is important that the tourists leave on projecting India as a modern country
India with a good impression in their for a comfortable holiday with a choice
minds so that they would, in turn, of beaches, historical attractions.
influence others to visit India. Factors,
which irritate the tourists, like REFERENCES
cumbersome immigration and customs 1. Kotler, Philip. Lane, Kevin. Marketing
Management. (Pearson Publications, 13th
procedures, unethical traders, difficulties Edition)
2. Malhotra, K Naresh. Marketing
in air or rail bookings for travel within Research, An Applied Orientation,
(Pearson Education, 2008)
India, unsanitary conditions at places of 3. Lovelock, Christopher. Wirtz, Jochen
stay and travel must therefore be given .Chatterjee, Jayanta. Services Marketing.(
Prentice Hall, 2002 ).
high priority for improvement. 4. Annual report of Ministry Tourism
Government of India 2008-2009.
The Marketing Strategy should 5. Annual report of Malaysia Tourism 2008
2009.
take into account the following factors: 6. World Travel & Tourism council report on
India and Malaysia Tourism.
The Asian tourism mart is growing at a 7. ASEAN Journal on Hospitality &
rapid rate. This makes the case for stiff Tourism, February 2009 - Chief Editor: Myra
P. Gunawan - Institut Teknologi Bandung,
competition among different nations to Bandung, Indonesia
8. Opportunities in Malaysian Tourism Industry
attract maximum tourists; Technical (2007-2009)
http://www.marketresearch.com/product/display.
assistance is required to draw asp?productid=1806312 August 25, 2008.
9.http://www.researchandmarkets.com/research/a
comprehensive master plans and also to 628d1/malaysia_tourism-
review and improve the existing plans; Authors: Rashid, Z.A., Mohamed Sharif Bashir
Author Affiliation: Department of Economics,
The competitive tourism trade warrants a Faculty of Economics and Management,
University Putra Malaysia, PM, 43400 Serdang,
constant and consistent marketing of Selangor, Malaysia.
10. Singh, Amrik.Outlook of Asia Pacific
new and exclusive destinations, which Tourism Industry: Current Trends and Future
(Singh, Amrik is an instructor, Department of
includes upgrading the existing ones; Parks, Recreation and Tourism, University of
For a number of years in the past, the Utah, Salt Lake City, Utah - email:
[email protected] )
emphasis has been on marketing India as 11.http://www.tourismstat.com/survey/public/sur
vey.
a cultural destination with rich and
diverse religious history and many
pilgrimage attractions. In the present
scenario, there is a need to diversify the
IJSER 2011
157
ISSN 2229-5518
Clinical and Computational Study of

Geometry & Heamodynamics of Arterial
Stenosis
Krittika Dasgupta, Abhirup Roy Choudhury, Abhijit Chanda, Debabrata Nag
Abstract Stenosis is abnormal narrowing of blood vessels. The presence of stenosis in arteries may cause critical flow
conditions. It may finally lead to stroke and heart-attack. A clinical study has been done on more than 130 patients along with
computational study using 2D axisymmetric rigid model of stenosis in the carotid artery. Assumed shapes of deposition zone
and the degree of occlusion used in the analysis were taken from clinical data. The Navier-Stokes equations for incompressible
fluid flow have been considered as the governing equations and it has been solved with varying flow parameters using standard
CFD software package. The radial velocity profiles at various points of the flow field, the centerline velocity plot and the
centerline pressure plots have been obtained from computational study and compared with the clinical data.
Index Terms Arterial flow, Clinical validation, Computational Fluid Dynamics, Heamodynamics, Mathemetical Modeling,
Stenosis, Stenosis geometry.
1 INTRODUCTION
I N the present century arterial stenosis is one of the

major causes behind death in all parts of the globe.
Arterial Stenosis is abnormal narrowing or restriction
elastic stenotic tube with various stenosis stiffness and
pressure conditions [4]. In past experiments blood flow
has been considered both as Newtonian or Non-
present in the inner wall of blood vessel due to the depo- Newtonian fluid depending upon the radius of the blood
sition of cholesterol, fatty materials, cellular waste etc. It vessel. Haemodynamic studies have been made for both
may happen in all large or small arteries, commonly in steady and pulsatile flows. However, no special emphasis
Coronary artery, Carotid artery, and Peripheral artery. In is given in the stenosis geometry and previous studies
the present case study we only emphasize on Carotid used idealized models using definite curves (Cosine
artery stenosis. Carotid artery is one of the larger arteries, curve, Smooth curves, Irregular Geometry). In this study,
present in our neck region. The normal geometry of the a detail care has been taken to obtain more realistic steno-
artery is divided into three segments, Common Carotid sis geometry after going through more than 130 patients
Artery (CCA), External Carotid Artery (ECA) & Internal real time Ultrasound Doppler Examination data.
Carotid Artery (ICA). The deposition of plaque may vary
in shape: simple to complex structures and in dimension. 2 CLINICAL STUDY
Flow through these complex structures is commonly 2.1 Data Collection
associated with flow separation, stagnation, recirculation,
More than 130 ultrasound images of vascular stenosis
secondary vortex motion, plaque rupture etc.
have been acquired for our analysis of Carotid arterial
Efforts have been made to model stenosis and its com-
stenosis. All these patient data have been collected ran-
plex hemodynamics by Computational Fluid Dynamics
domly from different well-known multi specialty hospit-
(CFD) and experimental analyses since 1990s. Ku and
als in eastern India. It has been ensured during data col-
others have made detailed studies on the fluid mechanics
lection and throughout the work that no patient identity
of vascular systems hemodynamic changes due to stenos-
is revealed. Only the information about age and sex has
es [1, 2]. Johnston and Kilpatrick (1991) studied the effect
been noted along with other necessary clinical informa-
of geometrical irregularities in the wall of a stenosed ar-
tion for every patient data.
tery for Reynolds numbers from 20 to 1000 [3]. Tang et al
[1995-1998] used 3D models for steady viscous flow in an
2.2 Study and Analysis

Each patient data are reviewed thoroughly and very care-
Krittika DasGupta: pursuing Masters Degree program in Bio-medical fully to identify the common geometry and occurrence of
Engineering Jadavpur University, India.
E-mail: [email protected] stenosis/plaque in the artery. Maximum and minimum
Abhirup Roy Choudhury: pursuing Bachelors Degree program in deposition heights, length of constriction, percentage of
Mechanical Eengineering in Jadavpur University, India. diametric reduction are also noted for each and every
E-mail: [email protected] data.
IJSER 2011
158
ISSN 2229-5518
3 COMPUTATIONAL STUDY 3.2 Mathematical Model

3.1 Physical Model 3.2.1 Governing equations
In the following study, an axis-symmetric geometry has The blood flow can be considered to be Newtonian when
been developed by considering the carotid artery to be a flowing through large arteries [1, 12]. In our study, as the
long straight pipe with radius R0=D0/2 and length common carotid artery has been dealt with the flow is
L0= 875R0, where D0 is taken as 5.7mm (validated from considered Newtonian, laminar, steady-state and incom-
different medical books). pressible.
The statistical analysis of the 130 patients reveals two The incompressible Navier-Stokes Equations along
dominant geometries with varied dimensions: with the continuity equation have been used as the
1. Curved shape (Fig 1) governing equation for modeling the fluid flow.
2. Rectangular shape (Fig 2)
(u.)u = .[-pI + ( u + ( u)T )] (1)
.u = 0 (2)
Where u is the axial velocity, p is the axial pressure,

is the dynamic viscosity and is the density of blood
and T is the transpose matrix. Equation is the momen-
tum balance equation and equation is the continuity
equation. In the current study the density of blood has
been considered as 1050 Kg/m3 and the dynamic viscosi-
ty as 0.00345 Pa.s.
Fig1.Model of curved stenosis used in study
3.2.2 Boundary conditions
The imposed boundary conditions are:
1. A fully developed velocity profile at the inlet. The
equation of the velocity profile is parabolic as ex-
pected in laminar flow
r 2
u (r ) u 1
R
where,
u(r)= radial velocity at an arbitrary radius
. Fig2.Model of rectangular stenosis used in study u =mean velocity
R = the radius of the artery
r = the radius at which the velocity is to be obtained.
2. A zero pressure with no viscous stress condition at
The occurrence of the deposition shows three domi- the outlet.
nant patterns 3. A no-slip condition at all the walls.
1. Single sided deposition i.e. u = 0
2. Axis-symmetric deposition
3. Non axis-symmetric deposition 3.3 Numerical Procedure
All of them are considered with a maximum diametric
Standard Finite Element CFD based software COMSOL
constriction of 62% which can be specified as a moderate
3.5a has been used for the solution of the problems. The
degree of stenosis, as a constriction of less than 50% is
solver type is parametric and the solver used is Direct
considered mild and above 70% is considered severe in
(PARDISO).
most medical literature.
Sufficient length of the artery downstream of the ste-
nosis has been taken so that the blood coming out of the 3.4 Mesh Details and Grid Sensitivity Test
constricted region is fully developed at the outlet of the A free mesh consisting of triangular elements has been
artery. The upstream length for all of the stenosis is con- used in the study with the maximum possible refinement.
sidered at Z=0.031. The mesh has been refined in the vicinity of the constric-
IJSER 2011
159
ISSN 2229-5518
tions so as to present a more accurate picture of their ef- indicate symmetric and asymmetric deposition.
fects on the blood flow. In the curved constriction, 14900
elements and for the rectangular constriction 15369 ele- TABLE2
ments have been used for solution of the problems. APPEARANCE OF PLAQUE
Grid sensitivity tests for all the simulations have been
performed. For all of the stenosis geometries, there have
been no noticeable changes in results when the grids have
been refined above the values mentioned. So the above
refinement of meshes is used in our subsequent studies.
3.5 Code Vallidation

In the absence of any standardized data regarding the
haemodynamics of stenosed arteries, results of flow
through a straight pipe without any constriction has vali-
dated the code. The plug fluid flow considered at the inlet
Four common shapes are seen which can be broadly
is fully developed after a certain distance from it. The
categorized as Cosine shaped, Rectangular shaped, conic-
radial velocity profile of the fully developed flow is para-
al shape and spherical or elliptical shaped. Apart from
bolic and the maximum velocity is the centerline velocity
this few irregular geometries are also observed. The
and its value is twice the mean velocity. All these results
common trend is towards the Cosine shaped geometry
are fully compliant with the known results of classical
(Fig.3) and rectangular geometry (Fig.4), but any possible
fluid mechanics.
combination of abovementioned geometry is noticed for
the deposition in both walls.
4 RESULTS
4.1 Clinical Results
Both of the carotid arteries have been viewed starting
from CCA to ICA and ECA the sites of deposition is as
follows:
TABLE1
AREA OF OCCURRENCE OF PLAQUE
Fig3. Cosine shaped stenosis
In this 2D longitudinal and cross sectional ultra sound

images plaque is visible in either side as well as both (in-
ner and outer) sides of the inner vessel wall. [Outer side
means upper wall boundary of the 2D ultrasound image.
Clinically it is towards periphery of the neck region and
the Inner side is the lower wall boundary of the 2D ul-
tra sound image. Clinically it is away from the periphery
of the neck region]. This type of both-sided plaque forma-
tion shows axis symmetric and non-axis symmetric for-
mations from single to multiple appearances.
Now, when we examine the 2D B-mode (black and
white mode) longitudinal section images for both side Fig4. Rectangular shaped stenosis
depositions more carefully, it shows maximum 18.51%
data with very small almost, non-measurable one sided Size variation is very much prominent through out the
deposition, where other side contributes for a good de- batch. Large size plaques are present along with multiple
gree of diametric reduction. Calcification is prominent in small size plaques. For the both (inner and outer) sided
25% cases for the over all batch. Cross sectional images deposition a comparison in length and width of geometry
IJSER 2011
160
ISSN 2229-5518
is provided in Fig 5 and Fig 6. 4.2.1Centerline pressure plot

From the centerline pressure plot, it has been observed
that at the inlet of the stenosis the pressure fall is higher
for higher values of Reynolds Number (Re). Even nega-
tive pressure values have been found in case of Re = 1000.
But at the outlet of the stenosis, the flows with higher Re
show higher values of pressure. So, the irreversible pres-
sure rise increases with increasing values of Re. (Fig. 7)
When comparing the pressure profiles of rectangular
and curved stenoses at a fixed Reynolds Number, the ir-
reversible pressure rise has been found to be higher in
case of the rectangular geometry. At Re = 1000, the pres-
sure rise for the rectangular stenosis is found to be 23%
higher than the curved geometry. (Fig. 8)
Fig5. Comparison of length of deposition in both walls
Fig6. Comparison of width of deposition in both walls

Fig7. The centerline pressure plots of a curved stenosis at different
Reynolds Numbers
From the Fig 5, it is clear that in all the cases of small

(<5mm), large (10-15mm) and very large (>15mm) depo-
sition, inner side plaque dimension is much more than
outer side. Even in case of maximum width of Deposition
the combined effect of both inner and outer wall maxi-
mum width provides a high degree of diametric reduc-
tion.
More stenosis is found in Male patients as compared
to Females falling in same age group, though the sample
volume of the clinical study was not so large. Data shows
a chance of stenosis is more for person crossing the age of
50-70 years depending on their foodhabbit, life stye and
past medical records.
4.2 Computational Results

Fig8. Centerline pressure plots of Rectangular and curved stenosis
In the available literatures, blood has been found to flow at Reynolds Number=1000
with Reynolds Number (Re) between 100 to 1000. So in
this study, for both the stenosis geometries, the flow has 4.2.2 Radial velocity field
been studied with Re = 100, 400, 800 and 1000. A zone of
As seen in Fig.9 the radial velocity field at the outlet of the
recirculation and an irreversible pressure rise have been
stenosis shows negative velocity and the maximum value
observed at the outlet of the stenosis for both rectangular
of the negative velocity is higher for higher values of Re.
and cosine model. The following points have been
(For the rectangular stenosis, at the stenosis inlet, the
observed by studying the simulated results of the rectan-
maximum velocity has been found to shift from the
gular and curved stenoses.
centerline).
IJSER 2011
161
ISSN 2229-5518
Fig9. The radial velocity distributions for a curved stenosis

at some distance from outlet of the stenosis for different Fig11. The centerline velocity plot of a curved stenosis for
Reynolds Number different Reynolds Numbers
The maximum velocity in case of rectangular profile is

higher than the curved geometry.
Fig12. The centerline velocity plots of rectangular and curved

stenoses at Reynolds Number=1000
In our current study we are concentrating on two main

single sided deposition geometry i.e. cosine shape and
Fig10. Radial velocity distributions of rectangular and rectangular shape. But in our clinical study we not only
curved stenoses at Reynolds Number = 1000 get single deposition we encountered multiple deposi-
tions near about 20% cases.
4.2.3 Centerline velocity field

From the centerline velocity plot (Fig 11), the maximum
velocity in the entire sub-domain has been found in a
zone near the outlet of the stenosis for all the values of Re
used. Also the length of reattachment of the flow after the
flow separation has been found to increase with increas-
ing values of Reynolds Number.
The reattachment length of the rectangular stenosis is
around 10% more than the curved stenosis, as can be con-
cluded from (Fig. 12).
4.2.3 Centerline velocity field Fig13. Recirculation length vs. Reynolds Number
The recirculation lengths of both the rectangular and
curved stenoses plotted against the respective Reynolds
The data set also reflects good amount of large sized
Numbers show an almost linear variation. From this
plaque accompanied by smaller one or multiple deposi-
graph (Fig 13) it can be seen very clearly that the recircu-
tions, where their dimension, placement, and appearance
lation lengths of the rectangular stenosis is higher than a
varies. Table 3 shows the distance between two adjacent
curved stenosis for the same value of Reynolds number.
IJSER 2011
162
ISSN 2229-5518
depositions, where almost no gap is considered for dis- re-attachment.

tance of < 1mm, small gap is considered for the distance From the comparison of the rectangular and curved
between 1mm 3mm, large gap is taken for the distance stenosis, it is inferred that as the extent of the recircula-
between 3mm 6mm, and above this all the distances are tion length is longer in case of the rectangular stenosis
considered as very large gap. than the curved one, the rectangular stenosis has a higher
tendency to propagate. This is expected from the shapes
of the stenoses. The rectangular stenosis presents itself as
TABLE3
a bluff body in the line of flow of blood while the curved
DISTANCE BETWEEN MULTIPLE STENOSIS stenosis is more streamlined. Also that the irreversible
pressure rise for the rectangular geometry is higher
proves that it is more severe than a curved stenosis of the
same maximum constriction due to causes already dis-
cussed.
It can also be predicted from the above study that as
depositions continue to occur downstream of a curved
stenosis, it will eventually develop into a rectangular ste-
nosis if enough time is available. So, the adverse effects of
From the clinical study it is found that multiple stenos- a stenosis essentially increase with time.
es often occur in reality. It causes much more complexity
in flow pattern. With such multiple depositions, size and
shape of recirculation zone is expected to suffer variation. ACKNOWLEDGMENTS
Numerical Modeling with such real life situations involv-
A great support has been provided by Dr. Manoranjan
ing multiple stensosis is being done currently and will be
Mahapatra, Kalinga Hospital, Bhubaneswar; Dr. Ashok
reported later
Moulik, and CMRI Hospital, Kolkata by providing clini-
cal ultrasound Doppler data and also PURSE Project,
5 INFERENCE DST, Govt of India for funding the project.
From the clinical study of more than 130 patients with
carotid artery stenosis we come to the common inferences REFERENCES
after detailed study of it, which are stated as follows. [1] D.N Ku., Blood flow in arteries. Ann. Rev. Fluid Mech.vol. 29, pp. 399-
More deposition is seen in ICA with respect to ECA 434, 1997.
proves the haemodynamic nature and construction of [2] D.M Wootton., D.N.Ku., Fluid mechanics of vascular systems, diseases,
bifurcation geometry of human Carotid artery. A good and thrombosis. Annu. Rev. Biomed. Eng. Vol.01, pp. 299-329, 1999.
amount of deposition is noted in the bulb and near bifur- [3] P.R. Johnston and D. Kilpatrick, Mathematical modeling of flow
cation zone conforms the fact of flow separation and through an irregular arterial stenosis. Journal of Biomechanics vol. 24,
stagnation and deposition of substances in this type of pp. 1069-1077, 1991.
regions. [4] H.I. Anderson, R. Halden, T. Glomsaker, Effects of surface irregularities
Lower wall deposition is maximum for both single on flow resistance in differently shaped arterial stenosis. Journal of
and both sided deposition stating stagnation of particles Biomechanics vol. 33, pp. 1257-1262, 2000.
in the lower surface of a flowing liquid for a pipe flow. [5] D. Tang, C.Yang, D.N Ku., A 3-D thin-wall model with fluid-structure
From the geometry and appearance of multiple plaque interaction for blood flow in carotid artery with symmetric and asym-
and its internal distance it is clear that a single deposition metric stenosis. Computers and Structures vol. 72, pp.357-377, 1999.
initiate another deposition in its locality an along time [6] C. Bertolotti, V. Deplano, Three-dimensional numerical simulation of
this depositions merges to form a more complex and large flow through stenosed coronary bypass. Journal of Biomechanics vol.
structure which is clinic ally more dangerous. 33, pp. 1011-1022. 1999.
In the present work the flow of blood through ste- [7] P.K.Mandal, An unsteady anaysis of Non-Newtonian blood flow
nosed arteries has been studied by considering blood to through tapered arteries with stenosis. International Journal of Non-
be a Newtonian fluid and the flow to be laminar by vary- Linear Mechanics vol. 40, pp. 151-164, 2005.
ing the Reynolds Number. [8] A. Yakhot., L. Grinberg, N.Nikitin, Modeling rough stenoses by an
For fixed stenosis geometry, as the pressure rise in- immersed-boundary method. Journal of Biomechanics vol. 38, pp. 1115-
creases with increasing Reynolds Number, the heart has 1127, 2005.
to supply even more pressure to overcome this adverse [9] C.A Taylor., J.D Humphrey., Open problems in computational vascular
pressure gradient. Thus the effort of the heart increases, biomechanics: Haemodynamics and arterial wall stenosis. Comput.
leading to angina (pain in the heart). Also as the recircula- Methods Appl. Mech. Engrg. Vol.198, 3514-3523, 2009.
tion zone is higher for higher values of the flow velocity, [10] COMSOL Multiphysics User Guide.
the tendency of the stenosis to propagate increases. This [11] COMSOL Multiphysics Modelling Guide.
is because the stenosis aggravates with higher lengths of [12] Jay D. Humphrey, Sherry L. Delange, An Introduction to Biomchanics.
IJSER 2011
163

ISSN 2229-5518
Segmentation Techniques for Iris

Recognition System
Surjeet Singh, Kulbir Singh
Abstract A biometric system provides automatic identification of an individual based on a unique feature or characteristic
possessed by the individual. Iris recognition is regarded as the most reliable and accurate biometric identification system
available. Iris recognition systems capture an image of an individual's eye, the iris in the image is then segm ented and
normalized for feature extraction process. The performance of iris recognition systems highly depends on segmentation and
normalization. This paper discusses the performance of segmentation techniques for iris recognition systems to increase the
overall accuracy.
Index TermsActive contour, Biometrics, Daugmans method, Hough Transform, Iris, Level Set method,
Segmentation.
system, known as Bertillonage, to fix the problem

1. INTRODUCTION of identification of convicted criminals.
R eliable personal recognition is critical to

many processes. Nowadays, modern
societies give higher relevance to systems that
In 1880, the British scientific journal Nature
published an article by Henry Faulds and
William James describing the uniqueness and
contribute to the increase of security and permanence of fingerprints. This motivated the
reliability, essentially due to terrorism and other design of the first elementary fingerprint
extremism or illegal acts. In this context, the use recognition system by Sir Francis Galton and
of biometric systems has been increasingly improved by Sir Edward R. Henry. Having
encouraged by public and private entities in quickly disseminated, the first fingerprint system
order to replace or improve traditional security in the United States was inaugurated by the New
systems. Basically, the aim is to establish an York State Prison Department in 1903 and the
identity based on who the person is, rather than first known convicted due to fingerprint
on what the person possesses or what the person evidences was reported in 1911.
remembers. Presently, due to increasing concerns
Biometrics can be regarded as the automated associated with security and the war on
measurement and enumeration of biological terrorism, biometrics has considerably increased
characteristics, in order to obtain a plausible its relevance. It has moved from a single and
quantitative value that, with high confidence, can almost standardized trait (fingerprint) to the use
distinguish between individuals. of more than ten distinct traits.
Although less automatized, biometrics has According to Matyas Jr. and Riha [1], every
been used - at least - for centuries. In the 14th biometric system depends on the features,
century, the Portuguese writer Joao de Barros whether genotypic or phenotypic it is based on.
reported its first known application. He wrote Similarly to Daugman [2], authors divide the
that Chinese merchants stamped childrens palm biometric traits into two types. Fried [3] and A.
print and footprints on paper with identification Bromba [4] classified the origin of the biometric
purposes. In the western world, until the late traits into three different types: genotypic,
1800s the automatic recognition of individuals behavioral, and randotypic.
was largely done using photographic memory. Following the proposal of Jain et al. [5],
In 1883, the French police and anthropologist biometric systems can be evaluated regarding
Alphonse Bertillon developed an anthropometric seven parameters: uniqueness, universality,
permanence, collectability, performance,
acceptability and circumvention.
Surjeet Singh is currently faculty Member at Faculty of Science & Figure 2 contains a comparison between the
Technology, The iCFAI University, Dehradun, Inida. E-mail: most common biometric traits. Each value was
[email protected]
Kulbir Singh is currenty Assistant Professor, in Electronics & obtained through averaging and weighting of the
Communication Engineering Department,Thapar University, Patiala, classifications proposed in [6], [7], [4], [8], [9],
India. E-mail: [email protected] [10] and [11].
For the purposes of our work, one of the most
important features is the ability to perform
IJSER 2011
164

ISSN 2229-5518
covert recognition, which can be performed by during embryologic development. The collarette
the fingerprint, face, iris and palmprint. Among divides the iris into the pupillary zone, which
these, iris must be enhanced, as it provides encircles the pupil, and the ciliary zone, which
higher uniqueness and circumvention values. extends from the collarette to the iris root. The
colour of these two zones often differs [12].
Genotypic Randotypic Behavioral The pupillary margin of the iris rests on the
anterior surface of the lens and, in profile, the iris
120%
100% has a truncated cone shape such that the
80% pupillary margin lies anterior to its peripheral
60% termination, the iris root. The root,
40% approximately 0.5 mm thick, is the thinnest part
20% of the iris and joins the iris to the anterior aspect
0% of the ciliary body. The iris divides the anterior
Palmprint
Fingerprint
Finger Geometry
Hand Geometry
Hand Vein
Signature
Voice
Keystroke
Iris
Face
segment of the globe into anterior and posterior
Retina
chambers, and the pupil allows the aqueous
humor to flow from the posterior into the
anterior chamber with no resistance.
Fig 1: Factors of influence of the biom etric traits.
Uniqueness Universatilty
Permanence Collectability
Performance Acceptability
Circumvention
120%
100%
80%
60%
40%
20%
0%
Finger
Hand
Iris
Fingerprint
Palmprint
Face
Keystroke
Signature
Voice
Hand Vein
Retina
Fig 2: Comparison between the most common biometric

traits (adapted and averaged from [6], [7], [4], [8], [9], [10]
and [11]). Fig 3: Surfaces and layers of the iris [12].
2. IRIS ANATOMY In figure 3 the iris cross-section shows the

The iris is a thin, circular structure located papillary (A) and ciliary portions (B), and the
anterior to the lens, often compared to a surface view shows a brown iris with its dense,
diaphragm of an optical system. The centre matted anterior border layer. Circular
aperture, the pupil, actually is located slightly contraction furrows are shown (arrows) in the
nasal and inferior to the iris centre. Pupil size ciliary portion of the iris. Fuchs crypts (c) are
regulates retinal illumination. The diameter can seen at either side of the collarette in the
vary from 1 mm to 9 mm depending on lighting pupillary and ciliary portion and peripherally
conditions. The pupil is very small (miotic) in near the iris root. The pigment ruff is seen at the
brightly lit conditions and fairly large (mydriatic) pupillary edge (d). The blue iris surface shows a
in dim illumination. The average diameter of the less dense anterior border layer and more
iris is 12 mm, and its thickness varies. It is prominent trabeculae. The iris vessels are shown
thickest in the region of the collarette, a circular beginning at the major arterial circle in the ciliary
ridge approximately 1.5 mm from the pupillary body (e). Radial branches of the arteries and
margin. This slightly raised jagged ridge was the veins extend toward the pupillary region. The
attachment site for the fetal pupillary membrane arteries form the incomplete minor arterial circle
IJSER 2011
165

ISSN 2229-5518
(f), from which branches extend toward the Based on the assumption that the pixels
pupil, forming capillary arcades. The sector intensity of the captured image can be well
below it demonstrates the circular arrangement represented by a mixture of three Gaussian
of the sphincter muscle (g) and the radial distributions, Kim et al. [28] proposed the use of
processes of the dilator muscle (h). The posterior Expectation Maximization [29] algorithm to
surface of the iris shows the radial contraction estimate the respective distribution parameters.
furrows (i) and the structural folds of Schwalbe They expected that Dark, Intermediate and
(j). Circular contraction folds also are present in Bright distributions contain the pixels
the ciliary portion. The pars plicata of the ciliary corresponding to the pupil, iris and reflections
body is at (k). [13] areas.
3.1 Daugmans Method
3. IRIS SEGMENTATION This is by far the most cited method [14] in the
In this paper CASIA iris image database has iris recognition literature. The author assumes
been used for the analysis of different both pupil and iris with circular form and
segmentation algorithms. CASIA iris image applies the following integrodifferential
database (version 1.0) includes 756 iris images operator:
from 108 eyes, hence 108 classes. I(x, y)
max , , G (r) ds (1)
In 1993, J. Daugman [14] presented one of the r , , 2r
most relevant methods, constituting the basis of This operator searches over the image
the majority of the functioning systems. domain (x, y)for the maximum in the blurred (by
Regarding the segmentation stage, this author a Gaussian Kernel G (r) partial derivative with
introduced an integrodifferential operator to find respect to increasing radius r, of the normalized
both the iris inner and outer borders. This contour integral of I(x, y) along a circular arc ds
operator remains actual and was proposed in of radius r and center coordinates (x , y ). In
2004 with minor differences by Nishino and other words, this method searches in the
Nayar [15]. space for the circumference center and radius
Similarly, Camus and Wildes [16] and with highest derivative values comparing to
Martin-Roche et al. [17] proposed circumferences of neighbour radius.
integrodifferential operators that search the At first the blurring factor is set for a coarse
space, with the objective of maximizing the scale of analysis so that only the very
equations that identify the iris borders. pronounced circular transition from iris to
Wildes [18] proposed iris segmentation (white) sclera is detected. Then after this strong
through a gradient based binary edge-map circular boundary is more precisely estimated, a
construction followed by circular Hough second search begins within the confined central
transform. This is the most common method, that interior of the located iris for the fainter pupillary
has been proposed with minor variants by Cui et boundary, using a finer convolution scale and a
al. [19], Huang et al.[20], Kong and Zhang[21], smaller search range defining the paths
Ma et al.[22], [23] and [24]. (x , y , r)contour integration. In the initial search
Liam et al. [25] proposed one interesting for the outer bounds of the iris, the angular arc of
method essentially due to its simplicity. This contour integration ds is restricted in range to
method is based in thresholds and in the two opposing 90 cones centered on the
maximization of a simple function, in order to horizontal meridian, since eyelids generally
obtain two ring parameters that correspond to obscure the upper and lower limbus of the iris.
iris inner and outer borders. Then in the subsequent interior search for the
Du et al. [26] proposed the iris detection pupillary boundary, the arc of contour
method based on the prior pupil segmentation. integration ds in operator (1) is restricted to the
The image is further transformed into polar upper 270 in order to avoid the corneal specular
coordinates and the iris outer border is detected reflection that is usually superimposed in the
as the largest horizontal edge resultant from lower 90 cone of the iris from the illuminator
Sobel filtering. However, this approach may fail located below the video camera. Taking the
in case of non-concentric iris and pupil, as well absolute value in (1) is not required when the
as for very dark iris textures. operator is used first to locate the outer
Morphologic operators were applied by Mira boundary of the iris, since the sclera is always
and Mayer [27] to obtain iris borders. They lighter than the iris and so the smoothed partial
detected the pupillary and scleric borders by derivative with increasing radius near the limbus
applying thresholding, image opening and is always positive. However, the pupil is not
closing. always darker than the iris, as in persons with
normal early cataract or significant back-
IJSER 2011
166

ISSN 2229-5518
scattered light from the lens and vitreous humor; Wildes et al. [18], Kong and Zhang [21], Tisse
applying the absolute value in makes the et al. [30] and Ma et al. [22] use Hough transform
operator a good circular edge-finder regardless to localize irises. The localization method, similar
of such polarity-reversing conditions. With to Daugman's method, is also based on the first
automatically tailored to the stage of search for derivative of the image. In the proposed method
both the pupil and limbus, and by making it by Wildes, an edge map of the image is first
correspondingly finer in successive iterations, the obtained by thresholding the magnitude of the
operator defined in has proven to be virtually image intensity gradient:
infallible in locating the visible inner and outer
annular boundaries of irises. Where and
. is a Gaussian
smoothing function with scaling parameter to
select the proper scale of edge analysis.
The edge map is then used in a voting process to
maximize the defined Hough transform for the
desired contour. Considering the obtained edge
points as , a Hough transform
can be written as:
where
Fig 4: Segmentation by Daugmans Method

For rapid discrete implementation of the
The limbus and pupil are both modeled as circles
integrodifferential operator in , it is more
and the parametric function is defined as:
efficient to interchange the order of convolution
and differentiation and to concatenate them,
before computing the discrete convolution of the
resulting operator with the discrete series of Assuming a circle with the center and
undersampled sums of pixels along circular radius , the edge points that are located over the
contours of increasing radius. Using the finite circle result in a zero value of the function. The
difference approximation to the derivative for a value of is then transformed to by the
discrete series in , function, which represents the local pattern of
the contour. The local patterns are then used in a
voting procedure using the Hough transform, ,
in order to locate the proper pupil and limbus
boundaries. In order to detect limbus, only
where is a small increment in radius, and vertical edge information is used. The upper and
replacing the convolution and contour integrals lower parts, which have the horizontal edge
with sums, we can derive through these information, are usually covered by the two
manipulations an efficient discrete operator (3) eyelids. The horizontal edge information is used
for finding the inner and outer boundaries of an for detecting the upper and lower eyelids, which
iris where is the angular sampling interval are modeled as parabolic arcs.
along the circular arcs, over which the summed We implemented this method in MATLAB
pixel intensities represent the contour by first employing Canny edge detection to
integrals expressed in . generate an edge map. Gradients were biased in
the vertical direction for the outer iris/sclera
3.2 Hough Transform boundary, as suggested by Wildes et al. [24].
Hough transform is a standard image analysis Vertical and horizontal gradients were weighted
tool for finding curves that can be defined in a equally for the inner iris/pupil boundary.
parametrical form such as lines, polynomials and The range of radius values to search for was
circles. The recognition of a global pattern is set manually, depending on the database used.
achieved using the local patterns. For instance, For the CASIA database, values of the iris radius
recognition of a circle can be achieved by range from 90 to 150 pixels, while the pupil
considering the strong edges in an image as the radius ranges from 28 to 75 pixels. In order to
local patterns and searching for the maximum make the circle detection process more efficient
value of a circular Hough transform. and accurate, the Hough transform for the
IJSER 2011
167

ISSN 2229-5518
iris/sclera boundary was performed first, then up curve evolution. Note that, when the function
the Hough transform for the iris/pupil boundary is constant , the energy functional in is the
was performed within the iris region, instead of area of the region . The
the whole eye region, since the pupil is always energy functional in can be viewed as
within the iris region. the weighted area of . The coefficient of
can be positive or negative, depending on the
relative position of the initial contour to the
object of interest. For example, if the initial
contours are placed outside the object, the
coefficient in the weighted area term should
take positive value, so that the contours can
shrink faster. If the initial contours are placed
inside the object, the coefficient should take
negative value to speed up the expansion of the
contours.
By calculus of variations , the Gateaux
derivative (first variation) of the functional in
can be written as
Fig 5: Segmentation by Hough Transform .

3.3 Variational Level Set Formulation of
Active Contours without Re-
initialization where is the Laplacian operator. Therefore, the
In image segmentation, active contours are function that minimizes this functional satisfies
dynamic curves that moves toward the object the Euler-Lagrange equation .The steepest
boundaries. To achieve this goal, we explicitly descent process for minimization of the
define an external energy that can move the zero functional is the following gradient flow:
level curve toward the object boundaries. Let be
an image, and be the edge indicator function
defined
this gradient flow is the evolution equation of the

where is the Gaussian kernel with standard level set function in the proposed method.
deviation . We define an external energy for a Left click to get iris points, right click to get end point
function as below 50 50
100 100
where and are constants, and the terms 150 150
and are defined by 200 200
250 250
50 100 150 200 250 300 50 100 150 200 250 300
200 iterations
Left click to get pupil points, right click to get end point
50 50
respectively, where is the univariate Dirac 100 100
function, and is the Heaviside function. Now,

150 150
we define the following total energy functional

200 200
The external energy drives the zero level set 250 250
toward the object boundaries, while the internal 50 100 150 200 250 300 50 100 150 200 250 300
energy penalizes the deviation of from a 50
signed distance function during its evolution. To 100
understand the geometric meaning of the energy,

150
we suppose that the zero level set of can
be represented by a differentiable parameterized 200
curve , . It is well known that the 250
energy functional ) computes the length of 50 100 150 200 250 300
Fig 6: (a) Selected iris mask (b) iris/sclera boundary (c)

the zero level curve of in the conformal metric
selected pupil mask (d) iris/pupil boundary (e) segmented
. The energy image (f) iris region.
functional in is introduced to speed
IJSER 2011
168

ISSN 2229-5518
The second and the third term in the right which could make the recognition process less
hand side of correspond to the gradient accurate, since there is less iris information.
flows of the energy functional and , However, this is preferred over including too
respectively, and are responsible of driving the much of the iris region, if there is a high chance it
zero level curve towards the object boundaries. would also include undetected eyelash and
To explain the effect of the first term, which is eyelid regions.
associated to the internal energy , we
notice that the gradient flow
has the factor as diffusion rate. If

, the diffusion rate is positive and the
effect of this term is the usual diffusion, i.e.
making more even and therefore reduce the
gradient . If , the term has effect of
reverse diffusion and therefore increase the
gradient [31].
4. RESULTS Fig 7: Illustrate the results of the Integrodifferential

The automatic segmentation model using operator over the pupils that are not perfect circles. The
circular contour does not detect pupil boundaries
Integrodifferential equations and Hough
accurately.
transform proved to be successful. The CASIA
database provided good segmentation, since
those eye images had been taken specifically for
iris recognition research and boundaries of iris
pupil and sclera were clearly distinguished. For
the CASIA database, the Hough transform based
segmentation technique managed to correctly
segment the iris region from 658 out of 756 eye
images, which corresponds to a success rate of
around 87% as compared to the Hough
transform based segmentation technique that
managed to correctly segment the iris region
from 624 out of 756 eye images, which
corresponds to a success rate of around 83%.
Using Integrodifferential equations and
Hough transform methods on locating the pupil Fig 8: Illustrate the results of the Hough transform operator
over the pupils that are not perfect circles. The circular
and limbus assume that the boundaries are
contour does not detect pupil boundaries accurately.
perfect circles. Although the approaches are 200 iterations 200 iterations
different, all these methods consider pupil and 50 50
limbus as circular curves. It has been noticed that

100 100
the circular assumption of the contours can lead

150
to inappropriate boundary detection Figure 7 150
and 8. 200 200
The above methods of segmentation resulted 250 250
in false detection due to noises such as strong 50 100 150 200 250 300 50 100 200 iterations 200
150 250 300
boundaries of upper and lower eyelids. The 50 50
strong eyelid boundaries and presence of

100
eyelashes affected the limbus localization 100
significantly. 150 150
We also implemented eyelashes and eyelids 200 200
detection for the above two methods. The eyelid 250 250
detection system proved quite successful, and 50 100 150 200 250 300 50 100 150 200 250 300
managed to isolate most occluding eyelid Fig 9: Illustrate the results of the active contour
regions. One problem was that it would segmentation method based on Level set evolution
without re-initialization over the pupils that are not perfect
sometimes isolate too much of the iris region,
circles.
IJSER 2011
169

ISSN 2229-5518
The eyelash detection system implemented segmentation results for the pupil and limbus
for the CASIA database also proved to be boundaries with success rate of almost 100%.
successful in isolating most of the eyelashes Only problem with this system was that the
occurring within the iris region as shown in initial contour was to be defined for each eye
Figure 11. A slight problem was that areas where image manually.
the eyelashes were light, such as at the tips were
not detected. However, these undetected areas REFERENCES
were small when compared with the size of the [1] V. Matyas and Z. Riha, Toward reliable user
iris region. authentication through biometrics, IEEE Security and
Privacy, vol. 1, no. 3, pp. 4549, 2003.
[2] J. G. Daugman, Phenotypic versus genotypic approaches
to face recognition, Face Recognition: From Theory to
Applications, pp. 108123. Heidelberg: Springer-Verlag,
1998.
[3] S. D. Fried, Domain access control systems and
methodology,http://www.itu.dk/courses/SIAS/E2005/AU22
40 01.pdf , 2004.
[4] M. Bromba, Biometrics FAQs,
http://www.bromba.com/faq/biofaqe.htm, 2010.
[5] A. K. Jain, R. Bolle, and S. Pankanti, Personal
Identification in networked society, 2nd edition. Kluwer
Academic Publisher, E.U.A., 1999.
[6] A. K. Jain, A.Ross, and S. Prabhakar, An introduction to
biometric recognition, IEEE Transactions on Circuits and
Systems for Video Technology, vol. 14, no. 1, pp. 419,
Fig 10: Automatic segmentation of image from CASIA January 2004.
database. Black region denote detected eyelid.
[7] S. Liu and M. Silverman, A practical guide to biometric
security technology, IT Professional, vol. 3, no 1, pp. 27
32, January 2001.
[8] Biometrics and the courts,
http://ctl.ncsc.dni.us/biomet%20web/BMIndex.html,
2010.
[9] Idesias Biometric Technologies. Biometric comparison
table,http://www.idesia-
biometrics.com/technology/biometric_comparison_table.
html, 2010.
[10] International Biometric Group, Which is the best
biometric technology?,
http://www.biometricgroup.com/reports/public/report
s/best_biometric.html, 2010.
[11] J. D. Woodward, K. W. Webb, E. M. Newton, M. A.
Bradley, D. Rubenson, K. Larson, J. Lilly, K. Smythe, B.
Fig 11: The eyelash detection technique, eyelash regions
are detected using thresholding and denoted as black. Houghton, H. A. Pincus, J. Schachter, and P. Steinberg,
Army Biometric Applications - Identifying and
TABLE I
Addressing Socio-Cultural Concerns, Rand Corporation,
COMPARISON OF DIFFERENT SEGMENTATION
Santa Monica, 2001.
TECHNIQUES.
[12] A. K. Khurana, Comprehensive Ophthalmology, New
Method No. Of Properly Accuracy Age International (P) Ltd., 4th edition, 2007.
eye Segmented [13] L. A. Remington, Clinical Anatomy of the Visual System,
images Elsevier Inc., 2nd edition, 2005.
Daugmans 756 658 87% [14] J. G. Daugman, High confidence visual recognition of
Method persons by a test of statistical independence, IEEE
Hough 756 624 83% Transactions on Pattern Analysis and Machine
Transform Intelligence, vol. 25, no. 11, pp. 11481161, November
Proposed 756 750 Approx: 1993.
Method 100% [15] K. Nishino and S. K. Nayar, Eyes for relighting, ACM
Trans. Graph., vol 23, no. 3, pp. 704711, 2004.
The proposed method of active contour [16] T.A. Camus and R. Wildes, Reliable and fast eye finding
segmentation based on Level set evolution in close-up images, Proceedings of the IEEE 16th
without re-initialization provided perfect
IJSER 2011
170

ISSN 2229-5518
International Conference on Pattern Recognition, pp. 389 Proceedings of the 25th International Conference on
394, Quebec, August 2002. Vision Interface, pp. 294299, Calgary, July 2002.
[17] D. Martin-Roche, C. Sanchez-Avila, and R. Sanchez- [31] C. Li, C. Xu, C. Gui, and M. D. Fox, Level Set Evolution
Reillo, Iris recognition for biometric identification using Without Re-initialization: A New Variational
dyadic wavelet transform zero-crossing, IEEE Aerospace Formulation, IEEE Computer Society Conference on
and Electronic Systems Magazine, Mag. 17, no. 10, pp. 3 Computer Vision and Pattern Recognition, vol. 1, pp. 430
6, 2002. 436, 2005.
[18] R. P. Wildes, Iris recognition: an emerging biometric
technology, Proceedings of the IEEE, vol. 85, no.9, pp.
13481363, U.S.A., September 1997.
[19] J. Cui, Y. Wang, T. Tan, L. Ma, and Z. Sun, A fast and
robust iris localization method based on texture
segmentation, Proceedings of the SPIE Defense and
Security Symposium, vol. 5404, pp. 401408, August 2004.
[20] J. Huang, Y. Wang, T. Tan, and J. Cui, A new iris
segmentation method for recognition, Proceedings of the
17th International Conference on Pattern Recognition
(ICPR), vol. 3, pp. 2326, 2004.
[21] W. K. Kong and D. Zhang, Accurate iris segmentation
method based on novel reflection and eyelash detection
model, Proceedings of the International Symposium on
Intelligent Multimedia, Video and Speech Processing, pp.
263266, Hong Kong, May 2001.
[22] L. Ma, Y. Wang, and T. Tan, Iris recognition using
circular symmetric filters, Proceedings of the 25th
International Conference on Pattern Recognition, vol. 2,
pp. 414417, Quebec, August 2002.
[23] L. Ma, T. Tan, Y. Wang, and D. Zhang, Personal
identification based on iris texture analysis, IEEE
Transactions on Pattern Analysis and Machine
Intelligence, vol. 25, no. 12, pp. 25192533, December
2003.
[24] L. Ma, Y. Wang, and D. Zhang, Efficient iris recognition
by characterizing key local variations, IEEE Transactions
on Image Processing, vol. 13, no. 6, pp. 739750, June
2004.
[25] L. Liam, A. Chekima, L. Fan, and J. Dargham, Iris
recognition using self organizing neural network,
Proceedings of the IEEE Student Conference on Research
and Developing Systems, pp. 169172, Malasya, June
2002.
[26] Y. Du, R. Ives, D. Etter, T. Welch, and C. Chang, A new
approach to iris pattern recognition, Proceedings of the
SPIE European Symposium on Optics/Photonics in
Defence and Security, vol. 5612, pp. 104116, October
2004.
[27] J. Mira and J. Mayer, Image feature extraction for
application of biometric identification of iris - a
morphological approach, Proceedings of the 16th
Brazilian Symposium on Computer Graphics and Image
Processing, pp. 391398, Brazil, October 2003.
[28] J. Kim, S. Cho, and J. Choi, Iris recognition using wavelet
features, Kluwer Academic Publishers, Journal of VLSI
Signal Processing, no. 38, pp. 147 256, November 2004.
[29] A. P. Dempster, N. Laird, and D. Rubin, Maximum
likelyhood from incomplete data via the EM algorithm,
Journal of the Royal Statistic Society, vol. 39, pp. 138,
1977.
[30] C. Tisse, L. Martin, L. Torres, and M. Robert, Person
identification technique using human iris recognition,
IJSER 2011
171

ISSN 2229-5518
Investigation of X-ray Plasmon Satellites in

Rare Earth Compounds
Ajay Vikram Singh, Dr. Sameer Sinha
Abstract- We have Investigate and study to X-ray Plasmon Satellites in Rare earth compounds ( La2CuO4 , Nd2CuO4 , Gd2CuO4 ,
PrNiSb2 , NdNiSb2 , Pr(OH)3 , Nd(OH)3 , Sm(OH)3 )
Keywords- Surface Plasmon Satellites, Relative Intensity & Energy Separation
INTRODUCTION
I N the characteristic X-ray Spectra, Diagram as well

as non Diagram lines are present. Those lines which
fit in the conventional energy level diagram are
source of such energy loss in the excitation of
collective plasma oscillations of the electrons in the
solid. This gives rise to a series of plasma peaks of
called Diagram lines. & those lines which do not fit in
decreasing magnitude spaced by energy p where p
the conventional energy level diagram are called non
is the frequency of plasma oscillation.
diagram lines. It is also known as Satellites or Second
Auger peaks are also broadened by small
order lines. Satellites are generally of weak intensity
energy losses suffered by the escaping electrons. This
lines & are found close to more intense parent line.
gives rise to a satellite on the low energy of the Auger
The satellites which are observed on higher energy
peak. Energy loss peaks have well defined energy with
side are called high energy satellites (HES) whereas
to primary energy.
those are observed on lower energy side are called
The involvement of Plasmon oscillation in the
lower energy satellites (LES). First Siegbahn &
X-ray emission or absorption spectra of solids has been
Stenstroem observed these satellites in the K-Spectra
widely studied during the last few decades and has
of element from Cr (24) to Ge (32) while coster
been recognized that the electron electron interaction
theraeus & Richtmyer in the L-Spectra of element
has played an important role.
from Cu (29) to Sb (51) & Hajlmar, Hindberg &
This Paper is devoted to Investigate and study
Hirsch in the M-Spectra of elements from Yb (70) to U to X-ray Plasmon Satellites in Rare earth compounds
(92). Several theories were proposed from time to time ( La2CuO4 , Nd2CuO4 , Gd2CuO4 , PrNiSb2 , NdNiSb2 ,
to explain the origin of these satellites. Out of these Pr(OH)3 , Nd(OH)3 , Sm(OH) 3 )
theories the plasmon theory is found to be the most According to Plasmon theory , if the valence
suitable theory especially for those satellites. electron , before filling the core vacancy , also excites
Plasmon theory was first proposed by Bohm a Plasmon ,then the energy p needed for the
& pines which are extended by Housten, Ferrel, excitation of Plasmon oscillation is taken from the
Noziers & Pines. According to this theory the low transiting valence electron so that the emitted radiation
energy plasmon satellites are emitted when valence will be derived off an energy p and a low energy
electron excites a plasmon during the annihilation of satellites will emitted whose sepration from the main
core hole conversely if Plasmon pre exists, its energy X-ray line will correspond to p . On the other hand if
add up to the energy of diagram line. the Plasmon pre exists , during the X-ray emission
process , then , on its decay it can give its energy to the
The radiation less reorganization of electronic transiting valence electron before it annihilates the core
shell of an atom is known as Auger effect. Auger vacancy . Thus the energy of emitted X-ray photon
satellites have also been observed by Korbar and will be higher than the main emission line and by an
Mehlhorn [1] Haynes et at. [2] Edward and Rudd [3]. amount p giving rise to high energy satellite .
Theoretical explanation for K series Auger spectrum
was given by Burhop and Asaad [4] using intermediate MATHEMATICAL CALCULATION
coupling. Later on more refined theory, using In order to confirm the involvement of Plasmon
relativistic and configuration interaction has been used in the emission of X-ray satellites the relative
by Listengarter [5] and Asaad [6] intensity of single Plasmon satellites must be
In Auger primary spectra, one can also calculated . In this process first we deal with
observe secondary electron peaks close to the primary mathematical details of canonical transformation
peaks are produced by incident electrons which have carried out over the model Hamiltonian of the
undergone well energy losses. The most common system . Thus the energy separation E of the
low and high energy Plasmon satellite from the
IJSER 2011
172

ISSN 2229-5518
corresponding main line should be equal to the parameter and = 0.47 rs1/2 in the place of = (1+l/L)-
1
quantum of Plasmon energy p which is given used by Pardee et. al.(14) . The equation (3) contains
by [10] a series of terms. The first term of the equation is
purely extrinsic, while second term is purely intrinsic.
The other terms are containing the relative
ev 1 contributions of both extrinsic and intrinsic. The
specialty of this formula is that each term alone or
Where Z = No.of unpaired electrons , = Specific simultaneously with other terms is able to give the
relative intensity. This formula also includes both the
gravity & = Molecular Weight
categories mentioned by Bradshaw and gives better
results as compared than traditional methods for
calculation of the relative intensity. Using the values of
This equation can be derived as given below .
, and rs in equation (4)
From the classical consideration , we get the
Using the equation (4), the author has for the
frequency of Plasmon oscillation as
first time, calculated the relative intensity of Rare
earth compounds ( La2CuO4 , Nd2CuO4 , Gd2CuO4 ,
2 PrNiSb2 , NdNiSb2 , Pr(OH)3 , Nd(OH)3 , Sm(OH)3 ) and
Hence the amount of energy given to Plasmon Our calculated and estimated values are in agreement
becomes with the calculated values of J. C. Parlebas et al. [ 15 ]
and A. Szytula , B. Penc, A. Jezierski [ 16 ]
Ep = p =
Reference
In this equation we can write n = 1. Korbar H. & Mehlhorn W.A. ; Phys. 191,
(1966) 217.
Where , Z and W are defined above and L is 2. Haynes S.K. & Velinsky, M & Velinsky
the Avogadro number .By putting the numerical L.J. ; Nucl. Phys. A99 (1967), 537.
value of constant , we get the Plasmon energy as 3. Rudd M.E. & Edward & Volz, D.J. ;
Phys Rev. 151, (1966), 28.
ev 3 4. Asaad, W.N. & Burhop E.H.S. ; Proc.
Our calculated values of E have been Phys. Soc., London 71, (1958), 369.
compared with the Scroccos experimental value. 5. Listengarten, M.A. ; Bull Acad. Sci.
And We have also calculated the relative U.S.S.R., Phys. Ser. 26 (1962), 182.
intensity of plasmon satellites, which is different 6. Asaad, W.N. ; Nucl. Phy. 66, (1965b),
in different processes. If the excitation of 494.
plasmon occurs during the transport of the 7. M.Scrocco in photoemission spectra of
electron through the solid, it is known as Pb.(II) halide; Phys. Rev. B25 (1982)
extrinsic process of plasmon excitation. The 1535-1540 .
plasmon can also be excited by another method 8. M.Scrocco , Satellites in X-ray Photo
known as intrinsic process. In this process, electron spectroscopy of insulator I 32
excitation of plasmon takes place simultaneously (1985) 1301-1306
with creation of a hole. Bradshaw et al have 9. M.Scrocco , Satellite in X-ray Photo
further divided core hole excitation into two electron spectroscopy of insulators II 32
classes, (1985) 1307-1310
1 - Where the number of slow electrons are 10. L.Marton , L.B.Lader and H.
conserved. Mendlowitz; Adv. Electronic and Electro
2 - Where the number of slow electrons are not Physics; edited by L.M arton Academic ,
conserved New York 7 (1955) , 225 .
11. Surendra poonia and S.N.Soni , Indian
The Author has calculated relative intensity in journal of pure and applied physics ,
both the cases with new modification in the light of vol.45, feb.2007 pp-119-126
Bradshaw [12] and Lengreth [13] work, which 12. A. M. Bradshaw, Cederbaurn S.L, Domeke
explains that not only intrinsic process but extrinsic W. & Krause Jour. Phys C: Solid State
process and their relative contribution may also Phys. 7, 4503, 1974
contribute in relative intensities. The combined effect 13 D. C. Lengreth, Phys. Rev. Letter, 26,
of intrinsic and extrinsic plasmon excitation intensity 1229, 1971
variation was suggested by Lengreth as: 14. W. J. Pardee, G.D. Mahan, D. E. Eastman,
R.A. Pollak, L. Ley, F.R. McFeely, S.P.
i= = n 4 Kowalczky and D.A. Shirely, Phys. Rev.
B, 11, 3614, 1975.
The value of is taken as = 0.12rs which is purely 15 J.C.Parlebas et al. , J.Phys. France 51 , (1990) ,
intrinsic, rs = (47.11/ ws) 2/3 is dimensionless 639-650
IJSER 2011
173

ISSN 2229-5518
16. A. Szytula , B. Penc, A. Jezierski, Materials

Science-Poland, Vol.26, No. 3, 2008
IJSER 2011
174

ISSN 2229-5518
Table -1 Energy sepration E of Rare Earth compounds
Authors Calculated Experimental value of

Compounds Z Plasmon Energy Energy Separation
S.No.
Separation (Es) Ref. [15,16]
1
PrNiSb2 1 443.12 9.34 4.18 3.9
2
NdNiSb2 1 447.28 9.43 4.18 4.8
3
La 2CuO4 4 405.33 8.55 8.36 8.73
4
Nd2CuO4 4 415.99 8.77 8.36 9.01
5
Gd2CuO4 4 284.75 6.00 8.36 8.73
6
Pr(OH)3 7 191.90 4.05 11.06 11.70
7
Nd(OH)3 6 195.23 4.12 10.24 10.60
8
Sm(OH)3 4 201.35 4.24 8.36 8.50
Table -2 Relative Intensity of Rare Earth compounds
Experimental
Author
value of
Alpha Beta Calculated Intensity
S.No. Compounds Es Rs Relative
() () Relative Assignment
Intensity
Intensity
Ref. [15,16]
1
PrNiSb2 2.96 6.33 1.182 0.759 1.1011719 1.08 +2/2
2
NdNiSb2 2.96 6.33 1.182 0.759 0.4184298 0.42 -2/2
3
La 2CuO4 5.91 3.99 0.938 0.47864 0.37111941 0.411 -2/2
4
Nd2CuO4 5.91 3.99 0.938 0.47864 0.37111941 0.363 -2/2
5
Gd2CuO4 5.91 3.99 0.938 0.47864 0.37111941 0.388 -2/2
6 3*(+0.1+2/2
Pr(OH)3 7.82 3.31 0.855 0.39719 1.73677471 1.7
+3/62)
7 3*(+0.1+2/2
Nd(OH)3 7.24 3.48 0.877 0.41813 1.83197714 2
+3/62)
8 3*(+0.1+2/2
Sm(OH)3 5.91 3.99 0.938 0.47864 2.12073651 2.5
+3/62)
1
175

ISSN 2229-5518
* Reader , Ganpat Sahai Post Graduate College , Sultanpur ,U.P. India

** Associate Professor, Rajarshi Rananjay Sinh Institute of Management & Technology, Amethi , CSJ Nagar, U.P. , India
**E-Mail ID [email protected]
2
176
ISSN 2229-5518
Improvement of Power System Stability by

Simultaneous AC-DC Power Transmission
T.Vijay Muni, T.Vinoditha, D.Kumar Swamy
Abstract This paper presents the concept of simultaneous ac-dc power transmission.Long extra high voltage (EHV) ac lines
cannot be loaded to their thermal limits due to this instability occurs in the power system.With the scheme proposed in this
paper,it is possible to load these lines very close to their thermal limits.The conductors are allowed to carry usual ac along dc
superimposed on it.The advantage of parallel ac-dc transmission for improvement of transient stability and dynamic stability and
dampout oscillations have been established.Simulation study is carried out in MATLAB software package.The results shows the
stability of power system when compared with only ac transmission.
Index Terms Extra high voltage (EHV) transmission, flexiable ac transmission system (FACTS), HVDC, MATLab,
simultaneous ac-dc transmission, Power System Stability, Transmission Efficeincy
1 INTRODUCTION
H VDC transmission lines in parallel with EHV ac

lines are recommended to improve transient and
dynamic stability as well as to damp out oscilla-
This paper presents a simple scheme of simul-
taneous EHV ac-dc power flow through the same
transmission line with an object to achieve the advan-
tions in power system. Long EHV ac lines can not be tages of parallel ac-dc transmission. Simultaneous ac-
loaded to its thermal limit to keep sufficient margin dc transmission may also claim advantages in some
against transient instability. But for optimum use of specific applications LV (low voltage) and MV (Me-
transmission lines here is a need to load EHV ac lines dium voltage) system.
close to their thermal limits by using flexible ac trans- The flexible ac transmission system (FACTS) con-
mission system (FACTS) components .Very fast control cepts, based on applying state-of-the-art power elec-
of SCRs in FACTS devices like state VAR system (SVS), tronic technology to existing ac transmission system,
controlled series capacitor (CSC), static phase shiftier improve stability to achieve power transmission close
(SPS) and controlled braking resistors oscillations as to its thermal limit. Another way to achieve the same
well as to control the voltage profile of the line by con- goal is simultaneous acdc power transmission in
trolling the total reactive power flow. Only the basic which the conductors are allowed to carry superim-
idea is proposed along with the feasibility study using posed dc current along with ac current. Ac and dc
elementary laboratory model. The main object is to power flow independently, and the added dc power
emphasize the possibility of simultaneous ac-dc flow does not cause any transient instability.
transmission with its inherent advantage of power
flow control improves stability and damps out oscilla-
tions in power system.
EHV ac line may be loaded to a very high val-
ue if the conductors are allowed to carry superimposed
dc current along with ac current. The added dc power
flow does not cause any transient instability.

T.Vijay Muni received Masters Degree in Power & Industrial Drives from
JNT University, Kakinada, India in 2010.Presently he working as Assis-
tant Professor in Electrical and Electronics Department,Sri Sarathi Insti-
tute of Engineering & Technology, Nuzvid,India.PH-09000055144.
T.Vinoditha is currently pursuing masters degree program in Electrical
Power Systems in JNT University, Hyderabad, India. PH-09052352600. 2 COCEPT OF SIMULTANEOUS AC-DC
D.Kumar Swamy recwived Masters Degree in EPSHV from JNT Universi-
TRANSMISSION
ty, Kakinada, India .Presently he working as Associate Professor & Head in The circuit diagram in Figure1 shows the basic
Electrical & Electronics Engineering Department,Dr.Paul Raj Engineering
College,Bhadrachalem, India.PH-09866653638. scheme for simultaneous ac-dc transmission. The dc
E-mail: [email protected] power is obtained through the rectifier bridge and in-
IJSER 2011
177
ISSN 2229-5518
jected to the neutral point of the zigzag connected sec- are neglected can be written as
ondary of sending end transformer, and again it is re-
converted to ac by the inverter bridge at the receiving Sending end voltage:
end. The inverter bridge is again connected to the neu- Vs = AVR + BIR (1)
tral of zigzag connected winding of the receiving end
transformer. Star connected primary windings in place Sending end current:
of delta-connected windings for the transformers may Is = CVR + DIR (2)
also be used for higher supply voltage. The single cir-
cuit transmission line carriers both 3 phase ac and dc Sending end power:
power. It is to be noted that a part of the total ac pow- Ps+ jQS = (- VS V*R)/B* + (D*/B*) Vs2 (3)
er at the sending end is converted into dc by the ter-
tiary winding of the transformer connected to rectified Receiving end power:
bridge. The same dc power is reconverted to ac at the PR+jQR = (VS* VR)/B* - (A*/B*)VR2 (4)
received end by the tertiary winding of the receiving The expressions for dc current and the dc power, when
end transformer connected to the inverter bridge. the ac resistive drop in the line and transformer are
Each conductor of the line carries one third of the total neglected,
dc current along with ac current Ia .The return path of
the dc current is through the ground. Zigzag con- Dc current:
nected winding is used at both ends to avoid satura- Id = (Vdrcos - Vdicos)/(Rer+(R/3) Rci) (5)
tion of transformer due to dc current flow. A high val- Power in inverter:
ue of reactor, X d is used to reduce harmonics in dc cur- Pdi = Vdi x Id (6)
rent. Power in rectifier:
In the absence of zero sequence and third harmon- Pdr = Vdr x Id (7)
ics or its multiple harmonic voltages, under normal
operating conditions, the ac current flow will be re- Where R is the line resistance per conductor, Rcr and Rci
stricted between the zigzag connected windings and commutating resistances, and, firing and extinction
the three conductors of the transmission line. Even the angles of rectifier and inverter respectively and Vdr and
presence of these components of voltages may only be Vdi are the maximum dc voltages of rectifier and inver-
able to produce negligible current through the ground ter side respectively. Values of Vdr and Vdi are 1.35
due to high of Xd. times line to line tertiary winding ac voltages of re-
Assuming the usual constant current control of rec- spective sides.
tifier and constant extinction angle control of inverter,
the equivalent circuit of the scheme under normal Reactive powers required by the converters are:
steady state operating condition is shown in Fig.2.
Qdi = Pdi tanI (8)
Qdr = Pdr tanr (9)
CosI = (cos + cos ( + i) )/2 (10)
Cosr = (cos + cos ( + r) )/2 (11)
Where I and r are commutation angles of inverter

and rectifier respectively and total active and reactive
powers at the two ends are
Pst = Ps + Pdr and Prt = PR + Pdi (12)
Qst = Qs + Qdr and Qrt = QR + Qdi (13)
Total transmission line loss is:
PL = (Ps + Pdr) (PR + Pdi) (14)
Ia being the rms ac current per conductor at any point
of the line, the total rms current per conductor be-
comes:
The dotted line in the figure shows the path of ac re- I = sqrt (Ia2 + (Id/3)2) and PL 3I2R (15)
turn current only. The ground carries the full dc cur- If the rated conductor current corresponding to its al-
lowable temperature rise is Ith and
rent Id only and each conductor of the line carries Id/3
Ia = X * Ith; X being less than unity, the dc current be-
along with the ac current per phase
The expressions for ac voltage and current and the comes:
Id = 3 x (sqrt (1-x2) ) Ith (16)
power equations in terms of A,B,C and D parameters
The total current I in any conductor is asym-
of each line when the resistive drop in transformer
metrical but two natural zero-crossings in each cycle in
winding and in the line conductors due to dc current
IJSER 2011
178
ISSN 2229-5518
current wave are obtained for (Id/3Ia) <1.414. cvts as used in EHV ac lines are used to measure ac
The instantaneous value of each conductor voltage component of transmission line voltage. Superimposed
with respect to ground becomes the dc voltage Vd with dc voltage in the transmission line does not affect the
a superimposed sinusoidally varying ac voltages hav- working of cvts. Linear couplers with high air-gap core
ing rms value Eph and the peak value being: may be employed for measurement of ac component of
Emax = V + 1.414 Eph line current as dc component of line current is not able
Electric field produced by any conductor vol- to saturate high air-gap cores.
tage possesses a dc component superimposed with Electric signal processing circuits may be used to
sinusoidally varying ac component. But the instanta- generate composite line voltage and current wave-
neous electric field polarity changes its sign twice in forms from the signals obtained for dc and ac compo-
cycle if (Vd/Eph) < 1.414.Therefore, higher creepage nents of voltage and current. Those signals are used for
distance requirement for insulator discs used for protection and control purposes.
HVDC lines are not required.
Each conductor is to be insulated for Emax but
the line to line voltage has no dc component and
3 SELECTION OF TRANSMISSION VOLTAGE
ELL(max) = 2.45 Eph.Therefore, conductor to conductor The instantaneous value of each conductor voltage
separation distance is determined only by rated ac vol- with respect to ground becomes more in case of simul-
tage of the line. taneous ac-dc transmission system by the amount of
Assuming Vd/Eph = k the dc voltage superimposed on ac and more discs are
Pdc/Pac (Vd * Id)/(3 * Eph * Ia * cos) = (k * sqrt(1- to be added in each string insulator to withstand this
x2))/(x * cos ) (17) increased dc voltage. However, there is no change re-
Total power quired in the conductor separation distance, as the
Pt = Pdc + Pac = (1 + [k * sqrt (1-x2)]/(x * cos)) * Pac (18) line-to-line voltage remains unaltered. Therefore,
tower structure does not need any modification if same
Detailed analysis of short current ac design of conductor is used.Another possibility could be that the
protective scheme, filter and instrumentation network original ac voltage of the transmission be reduced as
required for the proposed scheme is beyond the scope dc voltage is added such that peak voltage with re-
of present work, but preliminary qualitative analysis spect to ground remain unchanged. Therefore, there
presented below suggests that commonly used tech- would be no need to modify the towers and insulator
niques in HVDC/ac system may be adopted for this strings.
purposes.
In case of fault in the transmission system, gate signals 4 PROPOSED APPLICATIONS
to all the SCRs are blocked that to the bypass SCR s are
released to protect rectifier and inverter bridges. CBs 1.Long EHV ac lines can not be loaded to their ther-
are then tripped at both ends to isolate the complete mal limit to keep sufficient margin against transient
system. As mentioned earlier, if (Id3Ia) <1.414, CBs instability and to keep voltage regulation within al-
connected at the two ends of transmission line inter- lowable limit, the simultaneous power flow does not
rupt current at natural current zeroes and no special dc imposed any extra burden on stability of the system,
CB is required. To ensure proper operation of trans- rather it improves the stability. The resistive drop due
mission line CBs tripping signals to these CBs may to dc current being very small in comparison to im-
only be given after sensing the zero crossing of current pedance drop due to ac current, there is also no appre-
by zero crossing detectors. Else CBs connected to the ciable change in voltage regulation due to superim-
delta side of transformers (not shown in figure1) may posed dc current.
be used to isolate the fault. Saturation of transformer 2. Therefore one possible application of simultaneous
core, if any, due to asymmetric fault current reduces ac-dc transmission is to load the line close to its ther-
line side current but increases primary current of trans- mal limit by transmitting additional dc power. Figure3
former. Delta side CBs, designed to clear transformers shows the variation of Pt/Pac for changing values of k
terminal faults and winding faults, clear these faults and x at unity power factor. However, it is to be noted
easily. that additional conductor insulation is to be provided
Proper values of ac and dc filters as used in due to insertion of dc.
HVDC system may be connected to the delta side and 3. Necessity of additional dc power transmission will
zigzag neutral respectively to filter out higher harmon- be experienced maximum during peak load period
ics from dc and ac supplies. However, filters may be which is characterized with lower than rate voltage. If
omitted for low values of Vd and Id. dc power is injected during the peak loading period
At neutral terminals of zigzag winding dc cur- only with V d being in the range of 5% to 10% of E ph,
rent and voltages may be measured by adopting com- the same transmission line without having any en-
mon methods used in HVDC system. Conventional hanced insulation level may be allowed to be used For
IJSER 2011
179
ISSN 2229-5518
a value of x=0.7 and V d =0.05 E ph or 0.10 E ph, 5.1% or are used at each end. A supply of 3-phase, 400V, 50Hz
10.2% more power may be transmitted. are given at the sending end and a 3-phase, 400 V, 50
4.By adding a few more discs in insulator strings of Hz,1 HP induction motor in addition to a 3-phase,
each phase conductor with appropriate modifications 400V, 0.7 KW resistive load was connected at the re-
in cross-arms of towers insulation level between phase ceiving end. A 10 A, 110 Vdc reactor (Xd) was used at
to ground may be increased to a high value, which each end with the 230V zigzag connected neutral. Two
permits proportional increase in Emax, Therefore high- identical SCR bridges were used for rectifier and inver-
er value of Vd may be used to increase dc and total ter. The dc voltages of rectifier and inverter bridges
power flow through the line. This modification in the were adjusted between 145 V to135 V to vary dc cur-
exiting ac lines is justified due to high cost of a sepa- rent between 0 to 3A.
rate HVDC line. The same experiment was repeated by replacing the
5. With the very fast electronic control of firing angle rectifier at the sending and and the inverter at receiv-
( ) and extinction angle ( ) of the converters, the fast ing end by 24V battery and a 5A, 25 rheostat respec-
control of dc power may also be used to improve dy- tively, between Xd and ground.
namic stability and damping out oscillations in the The power transmission with and without dc com-
system similar to that of the ac-dc parallel transmission ponent was found to be satisfactory in all the cases. To
lines. check the saturation of zigzag connected transformer
6. Control of and also controls the rectifier and infor high value of Id, ac loads were disconnected and dc
verter VAR requirement and therefore, may be used to current was increased to 1.2 times the rated current for
control the voltage profile of the transmission line dur- a short time with the input transformer kept energized
ing low load condition and works as inductive shunt from 400V ac. But no changes in exciting current and
compensation. It may also be considered that the capa- terminal voltage of transformer were noticed verifying
citive VAR of the transmission line is supplying the no saturation even with high value of I d.
whole or part of the inductive VAR requirement of the
converter system. In pure HVDC system capacitance of
transmission line cannot be utilized to compensate
6 SIMULATION RESULTS
inductive VAR. The loadability of Moose (commercial name), ACSR,
7. The independent and fast control of active and reac- twinbundle conductor, 400-kV, 50-Hz, 450-km double
tive power associated with dc, superimposed with the circuit line has been computed.
normal ac active and reactive power may be consi-
dered to be working as another component of FACTS.
power upgrading by combining ac dc transmission
8. Simultaneous ac-dc power transmission may find its A+ a3
B+
application in some special cases of LV and MV distri- a3 A+
B+
C+
C+
A-
b3
b3 B-
bution system. c3
A-
B-
C- Di stributed Parameters Line
C- c3
Zigzag
When 3-phase power in addition to dc power is sup- Zigzag
Phase-Shi fting Transformer
Phase-Shifting Transformer1
plied to a location very near to a furnace or to a work

place having very high ambient temperature, rectifica- 500kV, 60 Hz 345kV, 50 Hz,
10,000 MVA equivalent
tion of 3-phase supply is not possible at that location 5000 MVA equivalent
i
A A A aA A + A Aa A A A
using semiconductor rectifier. In such place simultane- B B B bB B
+ -
Current Measurement Scope
+
B Bb B B B
ous ac-dc transmission is advantageous. C C C cC

Brect
C - - C Cc
Binv
C C C
phi = 80 deg. 3rd harm. Rectifier Inverter phi = 80 deg. 3rd harm.
In air craft 3-phase loads are generally fed with higher
frequency supply of about 400Hz and separate line is
used for dc loads. Skin effect restricts the optimum use a3 A+
A+
B+
a3
of distribution wires at high frequency. Simultaneous Open this block b3

B+
C+
A-
C+
A-
B-
b3
ac-dc power transmission reduces both volume and to visualize

recorded signals
c3
B-
C- Distributed Parameters Li ne1
C-
Zi gzag
c3
Zigzag
weight of distributors. Phase-Shifting Transformer2
Phase-Shifting Transformer3
A
B
C
Data Acquisition AC fil ters

50 Hz
9. Another possible application is the transmission of
A
B
C
AC fi lters 600 Mvar

60 Hz
dc power generated by PV solar cells directly to re- 600 Mvar Discrete,

Ts = s.
mote dc loads through 3-phase ac line. In all cases of Master Control
separate dc supply filter networks are not required. Rectifier

Control and Protection
Master Control
Inverter
Control and Protection
5 EXPERIMENTAL VERIFICATION
Fig 3: Simulink Model of Simultaneous AC-DC Trans-
mission
The feasibility of the basic scheme of simultaneous
ac-dc transmission was verified in the laboratory.
Transformer having a rating of 2 kVA, 400/230/110V
IJSER 2011
180
ISSN 2229-5518
Fig 6: Combined AC-DC currents

Fig 4 Sending end and receiving end voltages
TABLE I
COMPUTED RESULTS
Power Angle 30 45 60 75
AC Current(kA) 0.416 0.612 0.80 0.98
DC Current(kA) 5.25 5.07 4.80 4.50
AC Power(MW) 290 410 502 560
DC Power(MW) 1685 1625 1545 1150
Total Power(MW) 1970 2035 2047 1710
TABLE II
SIMULATION RESULTS
Power Angle 30 45 60 75
PS (MW) 2306 2370 2380 2342
Pac (MW) 295 410 495 540
Pdc (MW) 1715 1657 1585 1498
Pac loss (MW) 12 30 54 82
Pdc loss (MW) 280 265 241 217
PR (MW) 1988 2050 2060 1995
7 CONCLUSION
Fig 5: Sending and receiving currents A simple scheme of simultaneous EHV ac-dc
power transmission through the same transmission
line has been presented. Expressions of active and
reactive powers associated with ac and dc, conductor
voltage level and total power have been obtained for
IJSER 2011
181
ISSN 2229-5518
steady state normal operating condition. The possible

applications of the proposed scheme may be listed as:
loading a line close to its thermal limit, improvement
of transient and dynamic stability and damping of os-
cillations. In LV and MV distribution system the pro-
posed scheme may be applied in a workplace having
high ambient temperature or fed with high frequency
supply or with PV solar cells. Only the basic scheme
has been presented with qualitative assessment for its
implementation. Details of practical adaptation are
beyond the scope of the present work.
ACKNOWLEDGMENT
We are thankful to Department of Electrical
and Electonics Engineering of Sri Sarathi Institute of
Engineering and Technology, Nuzvid, India & Dr.Paul
Raj Engineering College, Bhadrachalem, India with
whom we had useful discussions regarding HVDC,
Performance of transmissions lines. Any suggestions
for further improvement of this topic are most wel-
come
REFERENCES
[1] N. G. Hingorani, FACTSflexible A.C. transmission system,
in Proc. Inst. Elect. Eng. 5th. Int. Conf. A.C. D.C. Power Trans-
mission,
[2] Padiyar.HVDC Power Transmission System. Wiley East-
ern, New Delhi, 1993)
[3] H. Rahman and B H Khan Stability Improvement of Power
Systemby Simultaneous AC-DC Power Transmission Electric
Power System Research Journal, Elsevier, Paper Editorial ID
No. EPSRD- 06-00732, Press Article No. EPSR-2560 Digital
Object.
[4] I W Kimbark.Direct Current Transmission Vol-I.Wiley,
New York, 1971.
IJSER 2011
182
ISSN 2229-5518
A Comprehensive Distributed Generation

Planning Optimization with Load Models
Mahmood Sheidaee, Mohsen Kalanter
Abstract In this paper a comprehensive model for Distribution Systems Planning (DSP) in the case of using Distributed
Generation (DG), with regard to load models is provided. Proposed model optimizes size and location of the distributed
generation. This model can optimize investment cost in distributed generation better than other solutions. It minimizes the
operating costs and total cost of the system losses. This Model affects the optimum location and size of the distributed
generation in distribution systems significantly. Simulation studies based on a new multiobjective evolutionary algorithm is
achieved. It is important that in the analysis made in this paper, DG is introduced as a key element in solving the DSP.
Moreover, the proposed method easily and with little development can satisfy the other goals.
Index Terms Economic Analysis, Distributed Generation, Distribution Systems Planning, Load Models.
1 INTRODUCTION
D istribution Companies (DISCOs) should apply new

strategies in order to increase economic power gen-
eration (because of load growth) and giving services
Similarly, in this paper, the effect of load models on DG
location and size planning optimization has been argued.
We will see that load models considerably affect on plan-
to customers and not being remained behind in market ning for location and size of DG within distribution net-
competition of electrical power. By using of new alterna- works. Also, an approach is given to solve the problem
tives, these goals are available for solution of planning which is based on Strength Pareto Evolutionary
problem of distribution system in addition to traditional Algorithm (SPEA).
ones.
The load growth value is predicted in traditional op-
tions so that it may reach to a certain level. Then, a new
2 DISTRIBUTION SYSTEM PLANNING MODEL
capacity shall be added to the given system. By consider- In the case of load growth in power electricity market,
ing new electrical substations or via expanding the capac- DISCO has two options to meet such demand.
ity of the existing substations through new feeder, such a 1- Scenario A: Purchasing the required extra power from
new capacity will be obtained by both of them [1], [2]. main grid and extending the existing substations in
One of the new options to planning for increase in capaci- distribution network. At this scenario, DISCO has to
ty is Distributed Generation. DG may lower total costs in develop the existing substations by installation of new
the system, decrease load flow within system, and im- transformers and upgrading some existing feeders
prove voltage profile [3], [4], and leading to decreased capacities if they have not sufficient thermal capacity, and
system losses [3]-[5], relieving the heavy loaded feeders purchasing power from the main grid.
and increase lifetime of equipments [6]. 2- Scenario B: Investment on DG as an alternative
From [7]-[10], a perfect revision is carried out on load candidate option for solving the DSP problem and
models which are applicable to load flow and dynamic purchase power from main grid and extending of the
studies. Such studies are conducted on frequency or vol-
existing substation.
tage dependent load models.
During recent years, the studies on evolutionary algo- A. Model Formulation
rithms have shown that these methods have not many of This paper aims to minimize the investment and
previous problems [11]. In general, these methods may operating costs of DG, reduction of active and reactive
obtain multiple pareto optimal solutions in one single losses, improvement in voltage profile and relieving the
run. heavy loaded feeders. It also conducts study about impact
This paper suggests using of DG by DISCOs, as a new of voltage dependent load models, namely, residential,
economic tool for Distribution System Planning (DSP) industrial and commercial load models within different
Problem. The proposed approach makes decision on DG scenarios of planning. Load models are defined as
optimal location and size and the optimized power which follows.
should inject through distribution system. The derived
results from this model may be used for bill estimation of
customers of Distribution Company. Two comprehensive Pi = P0i Vi / Qi = Q0i Vi (1)
scenarios will be discussed to cover various probabilities.
IJSER 2011
183
ISSN 2229-5518
where Pi and Qi are active and reactive power at bus i, P0i M T M

and Q0i are active and reactive power operating point in
bus I, Vi is voltage in bus i and and are active and
J Max
Cfi S DGi BK 8760 t Cri S DGi
i 1 t 1 i 1
reactive power exponents. In a constant power model (5)
T TN M V 2
conventionally used in power flow studied = = 0 is t ij
assumed. The values which are used for are active and
8760 Z ij
pf .Ce A B
t 1 i 1 j 1
reactive power exponents in industrial, residential and
commercial load models in this paper are given in Table 1 SS TU TN M T TU
[12]. A Ci,u Cij 8760 t
pf .Ce .Si,u (6)
TABLE 1 i 1 u 1 i 1 j 1 t 1 i 1
EXPONENT VALUES
Load Type TN M T TU
Constant 0 0 B Cij 8760 t
pf .Cint Sint .Sint Sint (7)
Industrial 0.18 6.00 i 1 j 1 t 1 i 1
Residential 0.92 4.04
Commercial 1.51 3.40 1
t (8)
During studying residential model, it is assumed that 1 d t
the system only has residential loads. Similarly, this is
also applied for industrial and commercial loads where where BK denotes backup capacity of DG (MVA), is
all these loads are of industrial and commercial types, present worth factor, Cf as investment cost in distributed
respectively. In real situations, loads arent exactly generations ($/MVA), Cr DG operating cost ($/MVA), Ce
residential, commercial and industrial, so the mixture is Electricity market price ($/MWh), Cij is Total feeder
load class should be foreseen for distribution system. cost from i to j ($), Ci,u as Potential transformer u in
There are several ideas to study on effect of DG within existing substation i cost ($), Cint Intertie electricity market
distribution systems. One of such ideas is the price ($/MWh), D is total load demand (MVA), d
computation of multiple indices to describe the effects of discount rate, i and j are bus indices, J as cost index ($), M
disperse generations on distribution system. These is total number of load buses, pf as system power factor,
indices are SDG is generation power of distributed generations
1) Active and Reactive Power Loss Indices (ILP and ILQ): (MVA), SDGMax denotes maximum capacity of distributed
generations (MVA), Sij power flow in feeder connecting
[PLDG ] [QLDG ] bus i to j (MVA), SijMax is feeder thermal capacity between
ILP = 100 / ILQ = (2)
[PL ] [QL ] 100 i and j buses (MVA), Si,u is dispatched power of
transformer u in substation i (MVA), Sint as amount of
power imported by the intertie (MVA), SSS is power
Where PLDG and QLDG are total loss of active and reactive purchased by the distribution utility (MVA), SSSMax is
power distribution system with DG, PL and QL are total capacity of existing substation (MVA), SS as number of
loss of active and reactive power of total system without existing substations, t incremental time intervals (year), T
DG in the distribution network. horizon planning year (year), TN total number of system
2) Voltage Profile Index (IVD): One of the advantage of buses, TU total number of substation transformers, V bus
proper location and size of the DG is the improvement in voltage (V), Vn system nominal voltage (V), V maximum
voltage profile. permissible voltage drop (V), and Zij feeder impedance
between buses i and j (ohms) [13].
V1 Vi Optimization should be minimized by consideration of
IVD max in 2 100 (3)
V1 various operational constraints. Such constraints are

given in (9) (14).
1) Total Power Conservation: By considering losses of
3) MVA Capacity Index (IC): This informational index
lines and power which is generated by DG (if it exists),
gives information in the field of system necessities for
sum of input and output powers should be equal to
promoting transmission line.
existing total load demand.
S ij
TN Vij2 M
IC max in2 (4)
CS ij
S ij
Z ij
S ji S DGj D j , j M (9)
i 1
i 1
4) Cost Index M$ (J):
IJSER 2011
184
ISSN 2229-5518
2) Distribution Feeders Thermal Capacity: Distribution Two three phase transformers with 10 MVA power ca-
systems feeders have a capacity limit for the total power pacity and 0.2 M$ price for each one may be also installed
flow through it. to expand the main substation. The cost of upgrading the
existing primary distribution feeder with another of high-
S ij S ijMax , i TN , j M (10) er capacity is 0.15 M$/Km. The price of other existing
equipments and feeders will be considered zero. System
power factor and discount rate are also 0.9 and 12.5%,
3) Distribution Substations Capacity: Power which is respectively. During all optimization runs, population
generated by substations shall be at the substations size and generations maximum number are 300 and 750
capacity level. respectively. The maximum size of Paretos optimal set
includes 20 solutions. The probabilities of crossover and
M mutation are 0.9 and 0.01, respectively. Recently, the stu-
Max
SSSij SSSi , i SS , j M (11) dies on evolutionary algorithms have indicated that these
algorithms may be effective in removal of problems of
j 1
4) Voltage Drop: The DISCO provides the older methods. The applied optimization algorithm is
predetermined maximum permissible voltage drop limit. SPEA [14].
0 Vi V j V , i TN , j M (12) 3 ANALYSIS AND RESULTS

Two argued main scenarios in this paper are simulated in
5) DG Operation: The generated power by DG shall be order to evaluate preference of investing on DGs in
lesser than DG capacity. solving DSP problems in comparison with other
traditional planning options.
Max
S DGi S DGi , i M (13)
3.1 Scenario A
6) Interties Delivery Power Capacity Limit: DISCO In this scenario, DG size is zero. This model gives the
determines the interties delivery power cost rates. Rate of optimal cost of substations new transformers, and the
delivered power depends on the purchased power value power which is dispatched by these transformers, and the
by distribution network. amount of delivered power from main grid to
distribution network for all types of load models (NL is
Cint S int 1.00Ce , S int 0,5MVA number of lines upgrading).
Results which obtained from optimization model are
1.05C e , S int 5,10MVA
(14) given in Table 2 for all load models. It is seen that load
1.15C e , S int 10,15MVA models affect on solutions. It is possible that the obtained
1.35C e , S int 15,20MVA solution does not apply to industrial load by using of
constant load model. Such an impact from load models is
also observed for residential, commercial, and composed
B. Primary Distribution System under Study
models. More money should be spent by purchase of new
The existing primary distribution system under this equipments and due to compensation for losses so these
study is a 9-bus system [13]. This system has a 40MVA costs are added to customers bills.
substation. Load growth which has been predicted for 4
years is approximately 28% (51.1 MVA). The given 3.2 Scenario B
system has 4 feeders with thermal capacity of 12 MVA.
Table 3 indicates the solutions that are derived for
Based on market indexes in 2002 (in USD), 70 $/MWh
investment on DG option. Comparing these results with
and 0.5 M$/MVA have been considered as prices of
the results of above scenario, we can see that investment
electricity market and natural gas generator set,
on DG presents a better planning. For constant load
respectively. DGs have a capacity which is multiples of 1
model, four groups of DG with capacities of 4, 4, 3 and 2
MVA at a generated electricity price of 50 $/MWh. An
MVA and 1 MVA have been selected as backup in buses
extra DG with 1 MVA capacity has been allocated for
7, 9, 3 and 9 in addition a single transaction taking place
each DG as backup [13]. It is assumed that DG has unity
by the intertie of 1.4927 MVA and the dispatched power
power factor [12]. The maximum capacity of DG has been
by expanding substation is zero (we do not expand the
provided for each bus (maximally 4 units plus one unit as
substation). Cost of planning for this load model is 2% at
backup). This is done in order to keep distributed
this scenario, active losses 52%, and the reactive losses is
generation concept and DG not to be concentrated as a
52% lesser than at first scenario. It is clear that by
centralized plant so the existing main substation to be
investing on DGs instead of purchase of power at higher
used at maximum level.
IJSER 2011
185
ISSN 2229-5518
prices, DISCO may reduce its consumers bill due to

decrease in cost. Table 3 shows detailed results which are
obtained for other load models in Scenario B.
DG, as a key element in DSP problem, is not used only
to minimize planning cost and reduced active and
reactive losses, but as it discussed, it also has several
economic, social and electrical advantages. Figures (1)
(5) show voltage profiles of distribution network buses
for all load models in both manners.
TABLE 2
RESULTS OF SCENARIO A Fig. 1. DISCOs buses voltage profile for constant load model
Index Constant Residential Industrial Commercial Mixture
J (M$) 31.3837 24.0529 26.7622 21.1386 25.0651
Sint
4.7820 2.9595 2.2405 4.9024 3.3959
(MVA)
Si,u
9.7132 7.6799 9.8189 4.1953 7.7625
(MVA)
PL
3.06 2.55 2.74 2.36 2.63
(MW)
QL
2.14 1.78 1.92 1.65 1.84
(MVAr)
NL 1 1 1 1 1
TABLE 3
RESULTS OF SCENARIO B
Index Constant Residential Industrial Commercial Mixture
J (M$) 30.7797 23.2452 26.0381 20.8633 24.7617
Sint Fig. 2. DISCOs buses voltage profile for industrial load model
1.4927 3.0012 0.889 4.7927 1.2087
(MVA)
Si,u
0 0 0 0.8567 0
(MVA)
PLDG
1.46 1.6 1.4 1.91 1.41
(MW)
QLDG
1.02 1.12 0.98 1.34 0.98
(MVAr)
SDG
2,3,4,4 4,4 2,4,2,3 4 3,2,4,1
(MVA)
DG
5,3,9,7 9,7 3,7,5,9 9 7,3,9,5
Location
NL 0 0 0 0 0
Voltage profiles are very better for all load models in

planning with DG than without it. Load models influence
on solutions.
One of the other advantages of introducing DG to solve Fig. 3. DISCOs buses voltage profile for residential load model
DSP problem has been shown in Figs (6) (10). As it seen,
feeders power flow is decreased for all load models, as a
result, system losses is reduced and at last, losses cost will
be decreased. Also, this reduces the feeders load and
subsequently, it increases feeders lifetime. Similarly, we
have a chance to use the existing distribution network
with no need to upgrading feeders for further load
growth. It is again observed that load models influence
on solutions. The solutions which are derived from
different load models do not apply to other models.
IJSER 2011
186
ISSN 2229-5518
Fig. 4. DISCOs buses voltage profile for commercial load model
Fig. 8. DISCOs primary distribution feeder power flow for residential

load model
Fig. 5. DISCOs buses voltage profile for mixture load model
Fig. 9. DISCOs primary distribution feeder power flow for

commercial load model
they are highly different from each other in cost values,

losses, voltage profile and feeders power flow, and this
indicates that for an appropriate location-size planning,
load models are important and crucial. As a new tool in
solving DSP problem in comparison with traditional
planning alternatives, investment on DGs may create
further economic, social advantages. As a result, DISCO
may restore its own customers and prevents these
customers not to buy electricity power from other
Fig. 6. DISCOs primary distribution feeder power flow for constant companies.
load model
Fig. 7. DISCOs primary distribution feeder power flow for industrial Fig. 10. DISCOs primary distribution feeder power flow for mixture
load model load model
IJSER 2011
187
ISSN 2229-5518
4 CONCLUSION [11] C. A. C. Coello, A comprehensive survey of evolutionary

based multiobjective optimization techniques, Knowledge and
A comprehensive analysis was presented for DG location- Information Systems, vol. 1, no. 3, 1999, pp. 269308.
size planning multiobjective optimization, including load [12] D. Singh, K. S. Verma, Multiobjective Optimization for DG
models, in distribution systems. It was seen that when Planning With Load Models, IEEE Trans. Power Syst., vol. 24,
no. 1, Feb. 2009, pp. 427-436.
load models are considered, some changes occur in DG [13] W.EI-Khattam, Y. G. Hegazy, M. M. A. Salama, An Integrated
location and size. To estimate DGs optimal location and Distributed Generation Optimization Model for Distribution
size, a new advanced optimization model was used. The System Planning, IEEE Trans. Power System, vol. 20, no. 2,
proposed optimization model adapts cost, system total May 2005, pp. 1158-1165.
losses, voltage profiles and feeders power flow and gives [14] M. A. Abido, Environmental/Economic Power Dispatch Using
Multiobjective Evolutionary Algorithms, IEEE Trans. Power
optimal answer which varied upon change of load Syst., vol. 18, no. 4, Nov.2003, pp. 15291537
models. The output results of this model not only give
DG location and size, but also express a need to cost in
order to purchase other new equipments (transformers
and feeders upgrading). Based on information and rates
that used at this paper, we observed that DG may lower
planning cost, improve voltage profile of distribution
system, decrease feeders power flow of distribution
networks, and minimize losses in distribution system and
increase their lifetime by lower down feeders load. Thus,
DG gives an opportunity to use the existing distribution
network to further load growth with no need to upgrade
feeders.
REFERENCES
[1] M. Ponnavaikko, K. S. Prakasa, and S. S. Venkata, Distribution
system planning through a quadratic mixed-integer program-
ming approach, IEEE Trans. Power Del., vol. PWRD-2, no. 4,
pp. 11571163, Oct. 1987.
[2] S. K. Khator and L. C. Leung, Power distribution planning: A
review of models and issues, IEEE Trans. Power Syst., vol. 12,
no. 3, pp. 11511159, Aug. 1997.
[3] P. P. Barker and R.W. De Mello, Determining the impact of
distributed generation on power systems. I. Radial distribution
systems, in IEEE Power Eng. Soc. Summer Meeting, vol. 3,
2000, pp. 16451656.
[4] X. Ding and A. A. Girgis, Optimal load shedding strategy in
power systems with distributed generation, in IEEE Winter
Meeting Power Eng. Soc., vol. 2, 2001, pp. 788793.
[5] N. Hadjsaid, J. F. Canard, and F. Dumas, Dispersed genera-
tion impact on distribution networks, IEEE Comput. Appl.
Power, vol. 12, no. 2, pp. 2228, Apr. 1999.
[6] L. Coles and R.W. Beck, Distributed generation can provide an
appropriate customer price response to help fix wholesale price
volatility, in IEEE Power Eng. Soc. Winter Meeting, vol. 1,
2001, pp. 141143.
[7] C. Concordia and S. Ihara, Load representation in power sys-
tems stability studies, IEEE Trans. Power App. Syst., vol. PAS-
101, no. 4, Apr. 1982, pp. 969977.
[8] IEEE Task Force on Load Representation for Dynamic Perfor-
mance, Bibliography on load models for power flow and dy-
namic performance simulation, IEEE Trans. Power Syst., vol.
10, no. 1, Feb. 1995, pp. 523538.
mance, Load representation for dynamic performance analy-
sis, IEEE Trans. Power Syst., vol. 8, no. 2, May 1993, pp. 472
482.
mance, Standard load models for power flow and dynamic
performance simulation, IEEE Trans. Power System, vol. 10,
no. 3, Aug. 1995, pp. 13021313.
IJSER 2011
188

ISSN 2229-5518
Positive Impact of Mobility Speed on

Performance of AODV at Increased Network
Load
Rajesh Deshmukh, Asha Ambhaikar
Abstract This paper studies impact of changing mobility speed on the performance of a reactive routing protocol AODV with
reference to varying network load. For experimental purposes, initially we observed the performance of AODV with increasing
Network Load from 4 packets to 24 packets at the maximum mobility speed of 10 m/s. In another scenario we observed the
performance of AODV with increasing Network Load from 4 packets to 24 packets at maximum mobility speed of 20 m/s. The
performance of AODV is observed across Packet Delivery Ratio, Loss Packet Ratio and Routing overhead parameters. Our
simulation results show that AODV is performing better with higher mobility speed at higher network load.
Index Terms AODV, MANET, Mobility Speed, Routing, Overhead, Random Waypoint
1. INTRODUCTION
A n adhoc network is a dynamic network. It allows

wireless mobile nodes dynamically forming a
temporary network without the use of any existing
tables). Ad-hoc routing protocols can be classified based
on different criteria. Depending upon the routing
mechanism employed by a given protocol, they fall in
network infrastructure or centralized administration. A two classes.
number of routing protocols like Dynamic Source Table Driven Routing Protocols (Proactive): Each node
Routing (DSR), Ad Hoc On-Demand Distance Vector in table-driven routing protocols, continuously
Routing (AODV) and Destination-Sequenced Distance- maintains up-to-date routes to every other node in the
Vector (DSDV) have been proposed. In this work an network. Periodic routing information is transmitted
attempt has been made to compare the performance of a throughout the network in order to maintain
reactive routing protocol for mobile ad hoc networks consistency of the routing table. Transmission occurs
AODV on the basis of varying number of packets with without delay if the route already exists, otherwise,
reference to mobility speed. The performance node needs to receive routing information
differentials are analyzed using varying mobility and corresponding to its destination while traffic packets
packet size. These simulations are carried out using the are waiting in the queue. Certain proactive routing
ns-2 network simulator, which is used to run ad hoc protocols are Destination- Sequenced Distance Vector
simulations. The results presented in this paper (DSDV), Wireless Routing Protocol (WRP), Global State
illustrate the importance in carefully evaluating and Routing (GSR) and Cluster head Gateway Switch
implementing routing protocols when evaluating an ad Routing (CGSR) [6].
hoc network protocol. On-Demand Routing Protocols (Reactive): In on
demand protocols, only when a node wants to send
packets to its destination it initiates a route discovery
1. AD HOC ROUTING PROTOCOLS process through the network. After a route is
Routing in Mobile Ad-hoc Network is a subject of determined or all possible permutations have been
extensive research, Because of the fact that it may be examined, the process of route discovery is completed.
necessary to pass several hops (multi-hop) before a The discovered route has to be maintained by a route
packet reaches the destination, a routing protocol is maintenance process until either the destination
needed. Routing protocol has two functions, first is becomes inaccessible along every path from the source
selection of routes for various source-destination pairs or until the route is no longer desired [6]. Some reactive
and second, Delivery of messages to their correct protocols are Cluster Based Routing Protocol (CBRP),
destination. Ad hoc On-Demand Distance Vector (AODV), Dynamic
The second function is conceptually straightforward Source Routing (DSR), Temporally Ordered Routing
using a variety of protocols and data structures (routing Algorithm (TORA), Associativity-Based Routing (ABR),
Signal Stability Routing (SSR) and Location Aided
IJSER 2011
189

ISSN 2229-5518
Routing (LAR) [6]. Routing (DSDV)

The Destination Sequenced Distance Vector Protocol
1.1 Dynamic Source Routing Protocol (DSR)
(DSDV) is a proactive, distance vector protocol which
The Dynamic Source Routing (DSR) protocol is an on uses the Bellmann - Ford algorithm [4]. DSDV is a hop
demand routing protocol based on source routing. DSR by hop distance vector routing protocol, wherein each
protocol is composed by two on-demand node maintains a routing table listing the next hop
mechanisms, which are requested only when two nodes and number of hops for each reachable destination.
want to communicate with each other. This Protocol is This protocol requires each mobile station to advertise,
composed of two essential parts of route discovery and to each of its current neighbors, its own routing table
route maintenance. Every node maintains a cache to (for instance, by broadcasting its entries). The entries in
store recently discovered paths [5]. Route Discovery this list may change fairly dynamically over time, so the
and Route Maintenance are built to behave according to advertisement must be made often enough to ensure
changes in the routes in use, adjusting them-selves that every mobile computer can almost always locate
when needed. Along with those mechanisms, DSR every other mobile computer of the collection. In
allows multiple routes to any destination, thus can lead addition, each mobile computer agrees to relay data
easily to load balancing or increase robustness .In the packets to other computers upon request. This
source routing technique, a sender determines the exact agreement places a premium on the ability to determine
sequence of nodes through which to propagate a packet. the shortest number of hops for a route to a destination
The list of intermediate nodes for routing is explicitly we would like to avoid unnecessarily disturbing mobile
contained in the packets header. In DSR [5], every hosts if they are in sleep mode. In this way a mobile
mobile node in the network needs to maintain a route computer may exchange data with any other mobile
cache where it caches source routes that it has learned. computer in the group even if the target of the data is
When a host wants to send a packet to some other host, not within range for direct communication. DSDV
it first checks its route cache for a source route to the requires a regular update of its routing tables, which
destination. In the case a route is found, the sender uses uses up battery power and a small amount of
this route to propagate the packet. Otherwise the source bandwidth even when the network is idle [4].
node initiates the route discovery process.
1.2 Ad hoc On-Demand Distance Vector (AODV) 2. MOBILITY MODEL
Ad hoc on demand distance vector (AODV) routing
protocol creates routes on-demand. In AODV, a route is
2.1 Random Waypoint Mobility Model
created only when requested by a network connection
and information regarding this route is stored only in The Random waypoint model is a random-based
the routing tables of those nodes that are present in the mobility model used in mobility management schemes
path of the route [1]. AODV is a reactive protocol based for mobile communication systems. Random Waypoint
upon the distance vector algorithm. The algorithm uses (RW) model assumes that each host is initially placed at
different types of messages to discover and maintain a random position within the simulation area [3]. The
links. Whenever a node wants to try and find a route to mobility model is designed to describe the movement
another node it broadcasts a Route Request (RREQ) to pattern of mobile users, and how their location, velocity
all its neighbors [2]. In this protocol, each terminal does and acceleration change over time [3]. Mobility models
not need to keep a view of the whole network or a route are used for simulation purposes when new network
to every other terminal. Nor does it need to periodically protocols are evaluated. In random based mobility
exchange route information with the neighbor simulation models, the mobile nodes move randomly
terminals. Furthermore, only when a mobile terminal and freely without restrictions. To be more specific, the
has packets to send to a destination does it need to destination, speed and direction are all chosen
discover and maintain a route to that destination randomly and independently of other nodes. This kind
terminal. In AODV, each terminal contains a route table of model has been used in many simulation studies.
for a destination [5]. A route table stores the following Two variants, the Random walk model and the Random
information: destination address and its sequence direction model are variants of the Random waypoint
number, active neighbors for the route, hop count to the model.
destination, and expiration time for the table. The In this model, a mobile node moves from its current
expiration time is updated each time the route is used. location to a randomly chosen new location within the
If this route has not been used for a specified period of simulation area, using a random speed uniformly
time, it is discarded [7]. distributed between [vmin, vmax] [3]. vmin refers to the
minimum speed of the simulation, vmax to the
1.3 Destination Sequenced Distance-Vector maximum speed [3]. The Random Waypoint Mobility
IJSER 2011
190

ISSN 2229-5518
Model includes pause times when a new direction and received by the destination through the number of
speed is selected. As soon as a mobile node arrives at packets originated by the CBR source.
the new destination, it pauses for a selected time period 2. Loss Packet Ratio (LPR) - Loss Packet Ratio is
(pause time) before starting traveling again. A Mobile calculated by dividing the number of packets that
node begins by staying in one location for a certain never reached the destination through the number
period of time (i.e. pause). Once this time expires, the of packets originated by the CBR source.
mobile node chooses a random destination in the
simulation area and a speed that is uniformly 3. Routing Overhead Routing overhead, which
measures the ratio of total routing packets sent and
distributed between [vmin, vmax]. The mobile node
the total number of packets sent.
then travels toward the newly chosen destination at the
selected speed. Upon arrival, the mobile node pauses
for a specified period of time starting the process again.
The random waypoint model is the most commonly 5. SIMULATION SETUP
used mobility model in the simulation of ad hoc In this simulation we wanted to investigate how
networks. It is known that the spatial distribution of mobility speed affects on the behavior AODV with
network nodes moving according to this model is non- increasing network load.
uniform. However, a closed-form expression of this TABLE 1
distribution and an in depth investigation is still
EVALUATION WITH MOBILITY SPEED 10 M/S
missing. This fact impairs the accuracy of the current
Parameter Value
simulation methodology of ad hoc networks and makes
Protocols AODV
it impossible to relate simulation based performance Simulation Time 100 s
results to corresponding analytical results. To overcome Number of Nodes 100
these problems, it is presented a detailed analytical Network Load 4, 8, 12, 16, 20, 24 Packets
study of the spatial node distribution generated by Pause Time 20 s
Environment Size 1020 m x 1020 m
random waypoint mobility. It is considered that a
Traffic Type Constant Bit Rate
generalization of the model in which the pause time of Maximum Speed 10 m / s
the mobile nodes is chosen arbitrarily in each waypoint Mobility Model Random Waypoint
and a fraction of nodes may remain static for the entire Network Simulator NS 2.33
simulation time [3]. TABLE 2
EVALUATION WITH MOBILITY SPEED 20 M/S
Parameter Value
3. THE TRAFFIC AND SCENARIO GENERATOR Protocols AODV
Simulation Time 100 s
Continuous bit rate (CBR) traffic sources are used. The
Number of Nodes 100
source-destination pairs are spread randomly over the Network Load 4, 8, 12, 16, 20, 24 Packets
network. The simulation uses Random Waypoint Pause Time 20 s
mobility model in a 1020 m x 1020 m field with varying Environment Size 1020 m x 1020 m
network load of 4 packets to 24 packets whereas Traffic Type Constant Bit Rate
Maximum Speed 20 m / s
mobility speed is kept at 10 m/s maximum. In the next
Mobility Model Random Waypoint
simulation network load is varied from 4 packets to 24 Network Simulator NS 2.33
packets, but this time mobility speed is kept 20 m/s
maximum. Here, each packet starts its journey from a
random location to a random destination with a 6. RESULTS AND DISCUSSIONS
randomly chosen speed. Once the destination is
reached, another random destination is targeted after a During the simulation we have increased the network
pause. The pause time, which affects the relative speeds load with maximum mobility maximum speed of 10
of the mobile hosts, is kept at 20s. Simulations are run m/s and recorded the performance of AODV. We did
for 100 simulated seconds. this simulation for 100 simulated seconds with
maximum 8 cbr connections. Readings were taken for
different network loads (4, 8, 12, 16, 20 and 24 packets).
4. PERFORMANCE METRICS Again same simulation is performed, but this time with
maximum speed of 20 m/s. From the results it is
Following important metrics are evaluated- evident that AODV starts to perform better with
1. Packet Delivery ratio (PDR) - Packet delivery ratio mobility speed of 20 m/s as compared to 10 m/s for
is calculated by dividing the number of packets same scenario. At higher network load and maximum
speed of 20 m/s, the Packet Delivery ratio increases,
IJSER 2011
191

ISSN 2229-5518
Loss Packet Ratio decreases and Routing Overhead

decreases.
7. PERFORMANCE EVALUATION
Observation for Mobility Speed of 10 m/s: Simulation
result in figure 1 shows that performance of AODV in
terms of Packet Delivery Ratio degrades as network
load is increased. When network load reach 12 packets,
PDR is dropped considerably. Even though PDR starts
to improve gradually from that point and reach a much
better performance around 16 packets of load. Once
again performance starts degrading, and continues to Fig 2. Number of Packets Vs Loss Packet Ratio
degrade more. Same with the Routing overhead, Figure
3 shows that Routing Overhead keeps on increasing
until network load of 12 packets and from that point
overhead starts to decrease till the network load reaches
16 packets. After this point routing overhead keeps on
increasing and never recovers again.
Observation for Mobility Speed of 20 m/s: Simulation
result in figure 1 shows that performance of AODV
degrades as network load is increased. A point to notice
is that when network load reach 12 packets,
performance of AODV is much improved as compared
to performance with Mobility Speed of 10 m/s. Packet
Delivery Ratio stays consistent until network load
reaches 16 packets, even though it is performing poor
than the earlier simulation scenario. PDR keeps on Fig 3. Number of Packets Vs Routing Overhead
decreasing until a point where network load reach 20
packets. From this point PDR starts to improve
gradually and achieves a much better performance as 8. CONCLUSION AND FUTURE WORK
compared to performance with mobility speed of 10 Empirical results illustrate that the performance of
m/s. About Routing overhead, Figure 3 shows that AODV varies widely across different network loads,
Routing Overhead remains either equal or better than and study results from two different scenarios shows
10 m/s scenario until network load reach 12 packets. that increasing the mobility speed does help to improve
Routing overhead stays consistent till 16 packets and the performance of AODV when it comes to higher
then again gets worse till the 16 packets mark. From network loads. Hence we have to consider the network
their AODV starts to improve the performance and load of an application while selecting the mobility
achieves better readings compared to reading with 10 speed.
m/s as network load crosses the 20 packets mark. The future scope is to find out what factors can bring
more improvements in performance of AODV not only
while the network load is further increased but also on
the load where AODV has not performed well in
simulations presented here. Further simulation needs to
be carried out for the performance evaluation with not
only increased mobility speed but also varying other
related parameters like Pause Time, Mobility models
etc.
REFERENCES
[1] Humaira Nishat, Vamsi Krishna K, Dr. D.Srinivasa Rao, Shakeel
Ahmed, Performance Evaluation of On Demand Routing
Protocols AODV and Modified AODV (R-AODV) in MANET,
Fig 1. Number of Packets Vs Packet Delivery Ratio International Journal of Distributed and Parallel Systems (IJDPS)
IJSER 2011
192

ISSN 2229-5518
Vol.2, No.1, January 2011, Engineering from Chhattisgarh Swami Vivekanand

http://airccse.org/journal/ijdps/papers/0111ijdps08.pdf Technical University, Bhilai. Currently he is working as
[2] Harminder S. Bindra, Sunil K. Maakar, A.L.Sangal, Performance Assistant Professor in Shri Shankaracharya Institute of
Evaluation of Two Reactive Routing Protocols of MANET using Professional Management & Technology (Department
Group Mobility Model, IJCSI International Journal of Computer of Computer Science Engineering), Raipur, India since
Science Issues, Vol. 7, Issue 3, No 10, May 2010, 2010. At present he is carrying out his research work in
http://www.ijcsi.org/papers/7-3-10-38-43.pdf Mobile Ad hoc Network Routing Protocols.
[3] Valentina Timcenko, Mirjana Stojanovic, Slavica Bostjancic Rakas,
MANET Routing Protocols vs. Mobility Models: Performance Asha Ambhaikar received B.E from Nagpur University,
Analysis and Comparison, Proceedings of the 9th WSEAS Nagpur, India, in Electronics Engineering in the year
International Conference on Applied Informatics and 2000 and later did her M.Tech in Information
Communications (AIC '09, http://www.wseas.us/e- Technology Allahabad Deemed University, India.
library/conferences/2009/moscow/AIC/AIC47.pdf Currently she is working as an Assistant Professor in
[4] Abdul Hadi Abd Rahman, Zuriati Ahmad Zukarnain, Rungta College of Engineering & Technology
Performance Comparison of AODV, DSDV and I-DSDV Routing
(Department of Information Technology), Bhilai, India.
Protocols in Mobile Ad Hoc Networks, European Journal of
She has published more than 14 research papers in
Scientific Research, ISSN 1450-216X Vol.31 No.4 (2009), pp.566-
reputed national and international journals &
conferences. Her area of interest includes, Computer
576, http://www.eurojournals.com/ejsr_31_4_07.pdf
Networking, Data warehousing and mining, Signal
[5] Sunil Taneja, Ashwani Kush, Amandeep Makkar, Experimental
processing, Image processing, and Information systems
Analysis of DSR, AODV using Speed and Pause time,
and Security.
International Journal of Innovation, Management and
Technology, Vol. 1, No. 5, December 2010, ISSN: 2010-0248,
http://www.ijimt.org/papers/79-M466.pdf
[6] Anuj K. Gupta, Dr. Harsh Sadawarti, Dr. Anil K. Verma,
Performance analysis of AODV, DSR & TORA Routing
Protocol, IACSIT International Journal of Engineering and
Technology, Vol.2, No.2, April 2010, ISSN: 1793-8236,
http://www.ijetch.org/papers/125--T203.pdf
[7] Parma Nand, S.C. Sharma, Rani Astya, Simulation Based
Parametric Analysis of AODV Protocol for Adhoc Network,
International Journal of Advanced Engineering & Applications,
Jan. 2010, http://steps-india.com/ijaea/2.pdf
ABOUT AUTHORS
Rajesh Deshmukh received B.Tech in

Information Technology Engineering
from Allahabad Deemed University in
the year 2007, and is final semester
student of M.Tech in Computer Science
IJSER 2011
193
ISSN 2229-5518
Development and Physical, Chemical and

Mechanical Characterization of Doped
Hydroxyapatite
Promita Bhattacharjee, Howa Begam, Abhijit Chanda
Abstract In this study, we made an attempt to synthesize doped bioactive hydroxyapatite (HAp) ceramic powder using a
simple Chemical method and studied its physical and mechanical properties. Different quantities (2wt% and 5 wt%) of
Magnesium chloride Hexahydrate , Zinc oxide, Titanium oxide were incorporated as dopants into Hap at the time of synthesis.
The synthesized powder samples were analyzed for their phases using X-ray diffraction technique, Fourier Transform Infrared
Spectroscopy. The synthesized powders were uniaxially compacted and then sintered at 1250C for 1hr in air. Vickers hardness
testing was performed to determine the hardness of the sintered structures. Fracture toughness of sintered samples was
calculated using Inverted Optical Microscope with Image Analysis software.
Index Terms Dopants, Fracture toughness, FTIR, Hydroxyapatite, Sintered structure, Vickers hardness, XRD.
1 INTRODUCTION
A mong different forms of calcium phosphates, the mineral during synthesis of HAp powder. This paper
bioactive hydroxyapatite (Ca10 (PO4)6(OH) 2) phase presents the synthesis and characterization of physical,
has been most extensively researched due to its out- mechanical and crystal structure of pure and doped HAp
standing biological responses to the physiological envi- ceramic in detail.
ronment. Hydroxyapatite is brittle in nature and load
bearing capacity and strength of HAp is low so we could
not use it in load bearing implant (total bone replace-
2 EXPERIMENTAL PROCEDURE
ment) where tensile stress is developed. To overcome 2.1 Materials and methods
the above stated limitations of Hap, many researchers
were tried to generate nano grained HAp powder [1, 2] in Pure and doped HAp powders were synthesized through
different methods. There is a significant difference of water based Chemical route method. In this method, Cal-
properties between natural and apatite crystals found cium hydroxide [Ca(OH)2](MERCK,INDIA) and Ortho-
in bone mineral and the conventional synthetic HAp. phosphoric Acid [H3(PO)4] (MERCK,INDIA) were used
Bone crystals are formed in a biological environment as raw materials to produce apatite particles .
through the process of biomineralization. In addition,
the bone mineral also contains trace ions like Na+, 10Ca (OH ) 2 6 H 3 PO4 Ca10 ( PO4 ) 6 (OH ) 2 18 H 2O
Mg++, K+, which are known to play a important role
in overall performance [1]. It has also been shown that The apatite powder produced was aged for 24 hr.
the bioactivity of conventional synthetic HAp ceramics is Then, the apatite particles in the suspension were fil-
inferior to the bone [1, 3-8]. During recent years, many trated, washed with ethanol three times, and dried at
researchers were tried in developing HAp powder 100C for 24 hr in air. The dried powder was ground
doped with metallic ions by Ball milling, dry and wet with a mortar and pestle into fine powder and sub-
milling process to increase the strength and ductility jected to calcinations at 800C temperature for 2 hrs
of HAp powders [1,3]. In this study, we have used a using electrically heated furnace ( NASKAR & Co.,
simple chemical route process that could produce HAp Model No.-EN170QT) at a constant heat rate of
powder with a fairly short synthesis time. We have 5C/min, followed by cooling inside the furnace. In
introduced three metal ions ,in different weight percen- order to synthesize HAp powder doped with Magne-
tage, which are known to be present as the bone sium, Zinc and Titanium, measured quantities of Mag-
nesium Chloride Hexahydrate (MgCl2,6H2O,MERCK,
INDIA, 96% pure), Zinc Oxide ( ZnO, MERCK, INDIA,
Promita Bhattacharjee, pursuing masters degree program in Biomedical 99% pure) and Titanium Oxide (TiO2, MERCK,INDIA,
Engineering in Jadavpur University, India. 98% pure) were incorporated into the Calcium Hy-
droxide suspension before the addition of H3(PO)4 solu-
Dr.Abhijit Chanda, joint director of School of Bioscience and Engineering tion separately. The dopants were used in the amount
Department of Jadavpur University,India, of 2wt% and 5 wt% to see their effects on powder
morphology and properties of the sintered ceramics.
IJSER 2011
194
ISSN 2229-5518
The resultant powders were then calcined at 800C

for 2 hr. in a electrically heated furnace. A constant 3.3.Powder Compaction And Sintering Study
heating rate of 5C/min. was used, followed by cool- As synthesized pure and doped crystalline HAp
ing inside the furnace. Various composition of doped powders were uniaxially pressed using a steel mold
HAp is presented in TableI. having an internal diameter of 12-13 mm at a pres-
TABLE I sure of 173MPa, with a 2-ton press for 2 min from
PEECO hydraulic pressing machine (PEECO Pvt Ltd,
ABBREVIATION,AMOUNT(WT%),COMPOSITIONOF ADDI-
TIVES AND PERCENTAGE OF ADDITION OF METALLIC IONS
M/C NO.-3/PR-2/HP-1/07-08). Green ceramic struc-
tures were measured for their density and then sin-
tered in a chamber furnace at 1250C for 1 hr at a con-
stant heating rate of 5C/min.
A sintering cycle was developed to achieve better
densification and to avoid cracking in the sintered
specimens, by introducing several soaking temperature
and tailoring the rate of heating and cooling.
Sintered ceramic structures were measured for their
density and then subjected to mechanical characteriza-
tion. Sintered samples intended for densification study,
hardness measurement. Green and sintered ceramic
specimens were measured for their geometric density
from the ratio between the mass of specimen and its
volume( determined by dimensional measurement).
3. POWDER CHARACTERIZATION
3.4. Mechanical Characterization
3.1 X-ray diffraction
X-ray diffraction (XRD) technique was used to study Hardness Test was carried out using a Vickers di-
the effect of calcination temperature and dopants on amond indenter on an automated hardness tester
the phase evolution and phase indentification. The (Model No-LV-700AT, LECO Co, MI). During the hard-
dried calcined powders at 800C were ground into ness test, a load of 0.3 Kgf, 1Kgf and 3Kgf was applied
fine powder using a mortar and pestle to break down on two samples of each of the composition type were
the powder agglomerates before analyzing in an X- tested for their hardness at three different locations
ray diffractometer. Powder samples were placed in the with three different loads. The average of these readings
specimen holder of a Rigaku diffractometer ( Model- were computed, reported and compared.
Miniflex, Rigaku Co., Tokyo, Japan) separately, and then To determine the Fracture Toughness, we used In-
analyzed, using Kb filtered Cu K radiation in the verted Optical Microscope (OLYMPUS Co. Ltd., Model
step scanning mode with tube voltage of 30KV and No-GX51F) interfacing with computer. We measured the
tube current of 15mA. The XRD patterns were record- average crack length (c) using Image Analysis Software
ed in the 2 range of 0-80 with scan speed .The cracks were developed by the indentation at the
1deg/min. The calcined powders doped with 2 wt% time of hardness testing. Fracture Toughness (K1c) was
and 5 wt% of [MgCl2, 6H2O] ZnO and TiO2, separately, calculated using simple equation considering radial-
were also analyzed for their phases in the same median crack geometry:-
manner. K1c=0.016(E/H) 1/2 P/(c) 3/2
Two samples of each of the composition type were meas-
3.2. Fourier transform infrared spectroscopy ured the fracture toughness.
Fourier Transform Infrared Spectroscopy (FTIR) relies
on the fact that most molecules absorb light in the 4. RESULT AND DISCUSSION
infrared region of the electromagnetic spectrum , this
absorption corresponds specifically to the bonds 4.1.Phase Identification
present in the molecule. In our experiment, we have
done FTIR of as prepared calcined HAp powder and
calcined HAp powder doped with different wt% of
[MgCl2, 6H2O] ZnO and TiO2.FTIR measurement were
performed in mid IR region (5000-400 cm-1) using
KBr pallets in a Perkin- Elmer, Model No- 1615 (USA)
instrument.
IJSER 2011
195
ISSN 2229-5518
4.1.1 XRD Analysis

X-ray diffraction data of the amorphous powder,
powder calcined at 800C temp. and Mg, Zn and Ti
doped HAp powder at different wt% calcined at 800
C ,were recorded in 2 range. The obtained XRD
patterns of pure HAp and HAp powder doped with
different wt% of Mg, Zn and Ti were presented in
Fig1.The pure HAp powder calcined at 800 C
exhibited several high intensity peaks corresponding
to various planes of HAp i.e. (0 0 2), (2 1 1), ( 1 1 2), (3 0
0) and ( 2 0 2) as revealed by our analysis, with
reference PDF card No.- 74-0566 for hydroxyapatite.
Compared to XRD pattern of pure HAp, c/a ratio of
HAp crystal structure and % of volumetric strain of
Figure 1(b). XRD patterns of Calcined HAp and Mg
HAp crystals (Table: I) changed due to addition of Mg,
doped HAp( 2wt% and 5wt%)
Zn and Ti. No phase change was noticed in their XRD
patterns.XRD figures 1(a),(b),(c),(d) shows the XRD
traces of the dried amorphous and calcined powders,
Calcined HAp Vs Zn doped HAp
with and without dopants. The XRD patterns of the
pure HAp and doped powders calcined at 800C clearly 700
show the presence of the most prominent peak at 2 600
angle of ~31.7C, corresponding to hydroxyapatite (2 1 In tensity (cps) 500
1) plane. Presence of this broad peak suggests that 400
Calcined HAp
2% Zn doped HAp
crystallites of HAp phase were formed as a result of 300
5% Zn doped HAp
calcinations at 800C. Almost identical patterns were 200
recorded for all these compositions, which suggest 100
that the presence of 2 wt% and 5 wt% of [MgCl2,6H2O] 0
ZnO and TiO2 as dopants did not alter the phase 0 20 40 60 80 100
purity of crystalline HAp. 2 theta(deg.)
Figure 1(c). XRD patterns of Calcined HAp and Zn

Figure 1(a). XRD patterns of Non Calcined HAp and

Non Calcined HAp
Figure 1(d). XRD patterns of Calcined HAp and Ti-
IJSER 2011
196
ISSN 2229-5518
Atomic radius of Mg (150 picometer),Zn (135 picome- C a lc in e d P u re H A p
ter) and Ti(144. 8 picometer) are different so they 1 20
changed c/a ratio by replacing cations and developed 1 00
volumetric strain.
80
TABLE II
C/A, UNIT CELL VOLUME AND PERCENTAGE OF VOLUMETRIC 60
%T
STRAIN
40
Composition c/a ratio of Unit cell vo- % Volume- 20
a lume tric
0
crystal (^3) strain
Pure HAp (before 0.731 529. 672 ____ 0 1 00 0 2 00 0 3 00 0 4 00 0 5 00 0
Calcination) W a v e n u m b e r(c m -1 )
Pure HAp( after 0.731 531. 697 _____
Calcination)
5% Mg doped HAp 0.734 533. 853 0. 789 %
(calcined) increase Figure 2(a). FTIR patterns of Calcined HAp
2% Mg doped 0.729 527. 725 0.368%
HAp decrease
(calcined)
5% Zn doped HAp 0.739 535. 059 1.017%
(calcined) increase
2% Zn doped HAp 0.733 531. 0451 0. 259 % 5wt% Mg doped HAp
(calcined) increase 120
2wt% Mg doped HAp
5% Ti doped HAp 0.720 526. 323 0. 632% Calcined HAp
(calcined) decrease 100

2% Ti doped HAp 0.720 526. 323 0. 632%
(calcined) decrease
80
4.1.2. FTIR ANALYSIS: 60

%T
The obtained FTIR patterns of pure HAp and HAp 40
powder doped with different wt% of Mg, Zn and Ti were

presented in Fig2.The Fig 2(a) shows that FTIR of pure 20
HAp powder calcined at 800C. In this FTIR plot of pure

0
calcined HAp powder we got bulge shaped peak at
3436 which reveals the presence of hydroxyl group. 0 1000 2000 3000 4000 5000
The Fig 2(b) shows that FTIR of 5wt% and 2 wt% Mg Wavenumber(cm-1)
doped HAp powder calcined at 800C. In these two

FTIR plots hydroxyl group peak is very short and tending
to disappear. This may due to the evaporation of the
surface moisture of the powder. The Fig 2(c) shows Figure 2(b). FTIR patterns of Calcined HAp and Mg
that FTIR of 5wt% and 2 wt% Zn doped HAp powder doped HAp( 2wt% and 5wt%)
calcined at 800C. In these two FTIR plots we got bulge
shaped peak at 3436 which shows the presence of
hydroxyl group. FTIR, spectra of HAp powder and ZnO
doped Hap powder present almost similar bands. The
Fig 2(d) shows that FTIR of 5wt% and 2 wt% Ti doped
HAp powder calcined at 800C.
IJSER 2011
197
ISSN 2229-5518
4.2.Sintering study and average of each of these compositions were calcu-

5 wt% Zn doped HAp
2 wt% Zn doped HAp
lated and presented in Table IV. Average diameter of sin-
140 Calcined HAp tered specimens is 10+_0. 05mm and thickness
120
4.5+_0.05mm. There was no major variation in shrinkage
among the various compositions. Results of our sinter-
100 ing study showed that sintering at 1250C helps in the
densification of pure and doped crystalline HAp ceram-
80
ics which is in line with literature on conventional
HAp sintering [5,6,10].We also performed some sinter-
%T
60
ing study at 1200C which showed relatively lower
40
densification(85. 25%).Sintering studies were not done
20
above 1300C, as it is already established that sintering
of HAp ceramics above 1300C leads to significant
0 phase change, unwanted grain growth and deteriora-
0 1000 2000 3000 4000 5000
tion of their properties. An average sintered density
Wavenumber(cm-1) of 2.9 g/cc were recorded in pure HAp specimens
sintered at 1250C which is equivalent to 93.9% of
theoretical density of HAp. In this work, presence of
2.(c) Mg, Zn and Ti separately as dopants during powder
synthesis again altered the sintered density of HAp
Figure 2(c). FTIR patterns of Calcined HAp and Zn powder. It is evident from the Table V that the Hap
doped HAp( 2wt% and 5wt%) doped with [MgCl2, 6H2O] and TiO2 did not show the
highest sintered density. Presence of 2% ZnO in the
HAp powder showed the best sintered density of 3.12
g/cc when sintered at 1250C.
5 wt% Ti doped HAp

2 wt% Ti doped HAp
Calcined HAp
120
TABLE III
100 GREEN DENSITY AND PERCENTAGE OF THEORETICAL DENSITY
80
Compositions Green densi- % of Theoreti-
ty(g/cc) cal
%T
60
density
40
Pure HAp(calcined) 1.66 52.41%
20
5%Mg doped HAP 1.69 53.36%
(calcined)
0 2%Mg doped HAP 1.597 50.43%
0 1000 2000 3000 4000 5000
(calcined)
Wavenumber(cm-1) 5%Zn doped HAP 1.64 51.78%
(calcined)
2%Zn doped HAP 1.66 52.42%
(calcined)
Figure 2(d). FTIR patterns of Calcined HAp and Ti 5% Ti doped HAP 1.65 52.01%
doped HAp( 2wt% and 5wt%) (calcined)
2%Ti doped HAP 1.56 49.26%
(calcined)
4.2.Sintering study
Green ceramic structure prepared via uniaxial
pressing were measured for their green density and
were subjected to pressureless sintering. Average
green density of all compositions are presented in
Table III .The average diameters of all specimens be-
fore sintering is 12.15+_0.05 mm and thickness 5.4+_0.05
mm. Each sintered specimen was measured for its
density and linear ,diametric and volumetric shrinkage
IJSER 2011
198
ISSN 2229-5518
TABLE IV higher than that with 5%.This feature may be attributed

SINTERED DENSITY,PERCENTAGE OF DIAMETRIC,LINEAR AND to higher density.
VOLUMETRIC SHRINKAGE Results of our hardness testing proved that the
hardness of crystalline HAp ceramics in influenced by
Composition Average % Diame- % Linear % Volume- the presence of Mg, Zn and Ti as dopants during syn-
Sintered tric Shrinkage tric thesis. As seen in graph of hardness in Fig 4 in 3 Kgf
Densi- Shrinkage Shrinkage
ty(g/cc) load we can present the Table VI.
Pure HAp 2.98 18.32% 18.72% 45. 8%
A2.0 2.87 18.27% 18.35% 45.34%
A5.0 2.87 16.64% 16.64% 43.45%

450
400
B 2.0 3.10 18.98% 18.61% 46.54% 350
H A R D N E S S (H V )
300
B5.0 2.85 17.31% 18.42% 44.15% PURE HAp
250
5% Mg
C2.0 2.89 18.87% 19.38% 46.94% 200
2% Mg
150
C5.0 2.78 16.54% 17.22% 42.31%
100
50
0
0.3 1 3
TABLE V
LOAD(Kgf)
PERCENTAGE OF CONVENTIONAL HAP DENSITY OF ALL COM-
POSITIONS
Composition % of conventional HAp density
Figure 3(a).Graphical representation of Hardness of
Pure HAp 93.9% Calcined HAp and Mg doped HAp( 2wt%
and 5wt%)
A 2. 0 90. 72%
A 5. 0 90. 52%
450
B 2. 0 97. 99% 400
350
B 5. 0 90. 096%
300
PURE HAp
C 2. 0 91. 14% 250
5%Zn
200
2%Zn
C 5. 0 87. 82%
150
100
50
4.3.MECHANICAL CHARACTERIZATION
0
4.3.1.VICKERS HARDNESS TESTING 0.3 1 3
The average hardness of each of these composition
LOAD(Kgf)
was calculated and plotted as a function of the dif-
ferent indentation loads (0.3Kgf, 1 Kgf and 3 Kgf) shown
in Fig 4. It is clear from the figures that the presence
of Mg, Zn and Ti dopants in crystalline HAp influence
Figure 3(b).Graphical representation of Hardness of
its hardness.
Calcined HAp and Zn doped HAp( 2wt% and 5wt%)
In almost each compositions 2wt% of doping caused
higher hardness values with highest load (3kgf).In other
loads also hardness for 2wt% dopant concentration was
3.(b)
IJSER 2011
199
ISSN 2229-5518
toughness with the Inverted Optical Microscope using

Image Analysis Software. The average fracture tough-
450 ness and their standard deviation of each of these
400 composition was calculated and presented in Table
350 VII. For radial-median crack system c/a ratio (where
c=average crack length, d= average diagonal) should be
300
250
PURE HAp more than 2. 5 and it was followed here.
5% Ti The average crack length was 271.18m for pure HAp,
200
2% Ti
150 225.05m for 2% Mg doped HAp, 171.56m for 5% Mg
100
doped HAp, 210.83m for 2% Zn doped HAp, 240.63m
50
for 2% Ti doped HAp and 82m for 5% Ti doped HAp.
0
We could not calculate the average crack length of 5% Zn
0.3 1 3 doped HAp because the crack propagation was not found
LOAD(Kgf) clearly from the site of indentation. The crack propaga-
tion path and the lateral crack growth of some speci-
mens during the time of indentation with 3 kgf load
are shown in Fig 4.
Figure 3(c).Graphical representation of Hardness of
Calcined HAp andTi doped HAp( 2wt% and TABLE VII
5wt%) AVERAGE FRACTURE TOUGHNESS AND STANDARD DEVIATION
Composition Average Frac- Standard

ture Toughness deviation
(MPam)
Pure HAp 0. 4988 0. 039823
TABLE VI
A 2. 0 0. 63552 0. 07592
AVERAGE HARDNESS AND PERCENTAGE OF HARDNESS VAR-
A 5. 0 0. 9993 ----
RIED
Composition Average % of Hardness in- B 2. 0 0. 7084 0.09215
Hardness crease/decrease B 5. 0 ----- -------
(HV) C 2. 0 0. 5973 0. 084834
C 5. 0 5.0735 0. 972112
Pure HAp 374. 31 __
A 2. 0 399.42 6.71% increase
A 5. 0 365.54 2.34% decrease
B 2. 0 399.71 6.78% increase
B 5. 0 320.593 14.35% decrease
C 2. 0 378.78 1.19% increase
C 5. 0 311.78 16.7% decrease
We can conclude from the above chart that percen-

tage of hardness increased 6.71%, 6.78% and 1.19% re-
spectively for 2 wt% of Mg, Zn and Ti dopants , but Fig. 4.(a). Vickers impressions and crack propagation
for 5 wt% of dopants of Mg, Zn and Ti % of hard-
in 5% Ti doped HAp taken by Optical Inverted Mi-
ness was decreased.
croscope
4.3.2.FRACTURE TOUGHNESS
Pure and doped crystalline HAp ceramics sin-
tered at 1250C were calculated for their fracture
We can conclude from the above chart that in most
IJSER 2011
200
ISSN 2229-5518
of the cases fracture toughness of are close to 1 or increased. For 5wt% of Ti doped HAp powder fracture
just less than 1. In the case of 5% Ti doped HAp( C 5. toughness was highly increased and has got clear impres-
0) fracture toughness was highly increased. Pure HAp sion with small cracks. In all compositions of doped HAp
powder mechanical property was increased, density was
high and average toughness values were increased. This
is marked improvement from practical point of view
compared with pure HAp powder.
6. Acknowledgment
We want to acknowledge Instrument Science and Engi-
neering Department, Mechanical engineering Department
of Jadavpur University for their technical help.
References
[1] ] S.J. Kalita, S. Bose, H.L. Hosick, A. Bandyopadhyay, Biomaterials 25

(2004) 2331.
[2] Yosuke Tanaka,Yoshihiro Hirata and Ryuichi Yoshinaka, Advanced

Nanostructure Material Science and Technology, 890-0065 ,Japan
Fig. 4.(b). Lateral crack propagation in Pure doped HAp
[3] S.J. Kalita, D. Rokusek, S. Bose, H.L. Hosick, A. Bandyopadhyay, J.
Biomed. Mater. SRes. A 71 (2004) 35.
F.C. Driessens, J.W. Van Dijk, J.M. Borggreven, Calcif. Tissue Res. 26
(1978) 127.
[4] R.Z. LeGeros, G. Bonel, R. Legros, Calcif. Tissue Res. 26 (1978) 111.
has inherent brittleness property. We can overcome this
brittleness property with these compositions. Fracture [5] M.A. Lopes, J.D. Santos, F.J. Monteiro, J.C. Knowles, J. Biomed.
Mater. Res. 39 (1998) 244.
toughness is increased for 2wt% and 5 wt% of all
compositions than pure HAp powder. It was also [6] R.A. Young, J. Dent. Res. 53 (1974) 193 (Suppl.).
noticed 5 wt% of all compositions is better than 2
wt% of all compositions in the case of fracture [7] S.J. Kalita, D. Rokusek, S. Bose, H.L. Hosick, A. Bandyopadhyay, J.
Biomed. Mater. SRes. A 71 (2004) 35.
toughness.
[8] G. Georgiou, J.C. Knowles, Biomaterials 22 (2001) 2811.
From the Fig 4(a) in case of 5% Ti doped HAp we have
[9] J.D. Santos, P.L. Silva, J.C. Knowles, S. Talal, F.J. Monteiro, J. Mater.
got very clear impression of rhombus with sharp edges. Sci., Mater. Med. 7 (1996) 187.
The impression resembles metal like impression with small
cracks at the edges. In contrary, pure HAp impression was [10] K.C.B. Yeong, J. Wang, S.C. Ng, Biomaterials 22 (2001) 2705.
not so clear and at identical load (3Kgf) it suffered severe
[11] M. Heughebaert, R.Z. LeGeros, M. Gineste, A. Guilhelm, G.Bonel,J.
chipping due to propagation and coalescence of lateral Biomed. Mater. Res. 22 (1988) 257.
cracks. It was also noticed that fracture toughness values of
5% Ti HAp showed large scatter. We got the lowest value
of fracture toughness of 5%Ti doped HAp was
1.4359MPam and highest value was 9.74MPam.
5. CONCLUSION
In our study, we prepared pure dense HAp and 2wt% .
and 5 wt% of Mg, Zn and Ti doped dense HAp powder
using chemical route method. In XRD study, all composi-
tions of doped and pure HAp powder we have got HAp
phase. Pure HAp and all compositions of doped HAp
powder we have got uniform pattern of shrinkage and
almost of all cases we have got densification above
90%. Hardness was increased for 2wt% of all composi-
tions of doped HAp powder; it may be attributed to bet-
ter densification. Pure HAp is brittle material, all compo-
sitions of doped HAp powder fracture toughness was
IJSER 2011
201
ISSN 2229-5518
A Chart for General and Complicated Data

Visualization
M Siluvairajah
Abstract Massive datasets arise naturally as a result of automated monitoring and transaction archival. Military intelligence data,
stock trades, retail purchases, medical and scientific observations, weather monitoring, spacecraft sensor data and censors data are
all examples of data streams continuously logged and stored in extremely large volumes, which create the need for innovative data vi-
sualization solutions. Although, there are many on-going researches and developments on this field recently, there are only few solu-
tions to visualize information for general public. In this paper, I explore different methods to use ManoStick chart to visualize informa-
tion.
Index Termsdata visualization, chart, graph, manostick
1 INTRODUCTION
There is a range of visualisation tools with a range of zation is its ability to visualize data, communicating in-
functions, such as; formation clearly and effectively. It doesnt mean that data
visualization needs to look boring to be functional or ex-
Comparison of values with bar charts, block his- tremely sophisticated to look beautiful. To convey ideas
tograms, bubble and matrix charts effectively, both aesthetic form and functionality need to
go hand in hand, providing insights into a rather sparse
Tag clouds to view word popularity in given text and complex data set by communicating its key-aspects in
Data point relationships with network diagrams a more intuitive way. Yet designers often tend to discard
and scatter plots the balance between design and function, creating gor-
geous data visualizations which fail to serve its main
Parts of the whole can be visualized with pie
purpose communicate information.
charts and tree charts
Track trend changes as rises and falls over time We can summarize that solutions for data visualization
with line graphs and stack graphs should be compact, based on known concept, easy to un-
derstand and able to express.
And many more different attempts to visualize
large scale information with complicated algo-
rithms to use by researches and analysts.
2 BACKGROUND
However, not all are suitable to current needs and prob- Introduction to ManoStick chart [4]: ManoStick chart has
lems at hand. five microstructures as shown on figure 1 & 2, which are
Some visualization methods such as bar charts, pie charts 1. Upper tail
and line graphs, are many decade old which had been 2. Upper body
created for the light-information on that time. 3. Middle body
4. Lower body
Some modern visualization [2] attempts such as Internet
Visualizations are too complicated. According to psychol- 5. Lower tail
ogist George Miller [3], who in a seminal paper in 1956,
noted the cognitive information bottleneck that percep- Each microstructure represents 20% of the depth or popu-
tion and attention span impose on the amount of informa- larity, which could be quantity when working with sales,
tion people are able to receive, process, and remember. volume when working with equities etc. ManoStick chart
In this otherwise serious essay, Miller suggested that sev- is better suitable for variable data with depth.
en independent items of information, plus or minus two,
might represent the limit of what most people can grasp To create a chart, we need to sort the variable data, find
in one moment. the total depth and divide the total depth with five as
there are five microstructures in Mano Stick.
According to Friedman [1], the main goal of data visuali-
IJSER 2011
202
ISSN 2229-5518
monthly, using ManoStick chart. Each ManoStick is to

represent the age of population for that month. By look-
ing at the graph, we can translate lot of information easi-
ly.
We can answer representative questions such as; what is

the oldest age? what is the mean average?, what age
group of people are highly populated? what age group of
people are least populated?
We can answer comparison questions such as; what is the

difference from oldest age between two months? What
age groups of people are growing from previous month?
Is the average age is high or lower than the previous
month?
Figure 1
In addition, it is possible to zoom microstructure in the
Now, first data set represents upper tail, the second data form of ManoStick for deep analysis. We can continue
set represents the upper body and so on. this process as long as there are data and analysis with
other soft of data such as their income in one group
To differentiate microstructures, we can use different co- against the other etc.
lours or patterns. Number five is convenient here as eve-
ryone used to it ever since the invention of analogue Further, we can divide ManoStick vertically to represent
clock. When the whole divided by five, we get 20% which to different set of data in this case, one side to represent
is each macrostructures weight. males and the other side to represent females.
Having four microstructures against one is rule of 80/20.

Italian economist Vilfredo Pareto created a mathematical
formula to describe the unequal distribution of wealth in 4 CONCLUSION
his country, observing that 20% of the people owned 80% We have seen that ManoStick charts are able to visualize
of the wealth. This opens up possibilities to analyse data the depth, can be easily learned even by general public
even with single stick. and take only small space to display. It can be used to
represent, compare and spot trends as well as expandable
with zooming. Now, it is possible to distribute more in-
formation easily to general public with a chart instead of
average house price, tomorrows temperature, cricket
players batting score, childrens exam result, people di-
agnosed with cancer or deceased people on the city last
month. It opens up many more opportunities not only for
statisticians but also for general public.
REFERENCES
[1] Vitaly Friedman (2008) "Data Visualization and Info graphics" in:
Graphics, Monday Inspiration, January 14th, 2008.
[2]http://www.webdesignerdepot.com/2009/06/50-great-
Figure 2
examples-of-data-visualization/
3. IMPLEMENTATION [3] Miller, George A. The Magical Number Seven, Plus or Minus
What sort of data visualization might help city planners Two: Some Limits on Our Capacity for Processing Information. The
to track their population? Think of a city like Paris where Psychological Review, 1956, vol. 63, pp. 81-97.
thousands of people move in, move out, babies born and
old people pass away. City planners are in constant chal- [4] M Siluvairajah, ManoStick: An Ariel View of the Stock Market,
lenge to offer amenities, education, health care, transport 2009, pp.84-91 ISBN 978-0956395603
and etc. while businesses need to supply the necessary
products and services. [5] Review of Visualization in the Social Sciences: A State of the Art
Survey and Report by Scott Orford, Daniel Dorling and Richard
Harris, 1998.
Here, we can create a time series graph, example
IJSER 2011
203
ISSN 2229-5518
Application of Reliability Analysis:

A Technical Survey
Dr. Anju Khandelwal
Abstract The objective of this paper is to present a survey of recent research work of high quality that deal with reliability in
different fields of engineering and physical sciences. This paper covers several important areas of reliability, significant research
efforts being made all over the world. The survey provides insight into past, current and future trends of reliability in different
fields of Engineering, Technology and medical sciences with applications with specific problems.
Index Terms CCN, Coherent Systems, Distributed Computing Systems, Grid Computing, Nanotechnology, Network
Reliability, Reliability.
1 INTRODUCTION
T HIS Traditional system-reliability measures include

reliability, availability, and interval availability. Re-
liability is the probability that a system operates
agreements? And, finally, how can one decide which ar-
tificial heart to place into a patient, given neither has ever
been inside a human before? To resolve this dilemma,
without interruption during an interval of interest under reliability must be: 1) reinterpreted, and then 2) quanti-
specified conditions. Reliability can be extended to in- fied. Using the scientific method, researchers use evi-
clude several levels of system performance. A first per- dence to determine the probability of success or failure.
formance-oriented extension of reliability is to replace a Therefore, reliability can be seen as an image of probabili-
single acceptable-level-of-operation by a set of perfor- ty. The redefined concept of reliability incorporates aux-
mance-levels. This approach is used for evaluating net- iliary sources of data, such as expert knowledge, corpo-
work performance and reliability. The performance-level rate memory, and mathematical modeling and simula-
is based on metrics derived from an application- tion. By combining both types of data, reliability assess-
dependent performance model. For example, the perfor- ment is ready to enter the 21st century. Thus, reliability is
mance-level might be the rate of job completion, the re- a quantified measure of uncertainty about a particular
sponse time, or the number of jobs completed in a given type of event (or events). Reliability can also be seen as a
time-interval. Availability is the probability that the sys- probability.
tem is in an operational state at the time of interest.
Availability can be computed by summing the state prob-
abilities of the operational states. Reliability is the proba-
2 APPLICATIONS OF RELIABILITY IN DIFFERENT
bility that the system stays in an operational state FIELDS
throughout an interval. In a system without repair, relia- Reliability is a charged word guaranteed to get attention
bility and availability are easily related. In a system with at its mere mention. Bringing with it a host of connota-
repair, if any repair transitions that leave failed states are tions, reliability, and in particular its appraisal faces a
deleted, making failure states absorbing states, reliability critical dilemma at the dawn of a new century. Tradition-
can be computed using the same methods as availability. al reliability assessment consists of various real-world
Interval availability is the fraction of time the system assessments driven by the scientific method; i.e., conduct-
spends in an operational state during an interval of inter- ing extensive real-world tests over extensive time periods
est. The mean interval availability can be computed by (often years) enabled scientists to determine a products
determining the mean time the system spends in opera- reliability under a host of specific conditions. In this 21st
tional states. Mean interval availability is a cumulative century, humanitys technology advances walk hand in
measure that depends on the cumulative amount of time hand with myriad testing constraints, such as political
spent in a state. and societal principles, economic and time considerations,
For example, how can traditional reliability assessment and lack of scientific and technology knowledge. Because
techniques determine the dependability of manned space of these constraints, the accuracy and efficiency of tradi-
vehicle designed to explore Mars, given that humanity tional methods of reliability assessment become much
has yet to venture that far into space? How can one de- more questionable. Applications are the important part of
termine the reliability of a nuclear weapon, given that the research. Any theory has importance, if it is useful and
world has in place test-ban treaties and international applicable. Many researchers are busy these days apply-
IJSER 2011
204
ISSN 2229-5518
ing concepts of Reliability in various fields of Engineering Network analysis is also an important approach to model
and Sciences. Some important applications are given here: real-world systems. System reliability and system unre-
liability are two related performance indices useful to
2.1 Nano-Technology measure the quality level of a supply-demand system. For
Nano-reliability measures the ability of a nano-scaled a binary-state network without flow, the system unrelia-
product to perform its intended functionality. At the nano bility is the probability that the system can not connect
scale, the physical, chemical, and biological properties of the source and the sink. Extending to a limited-flow net-
materials differ in fundamental, valuable ways from the work in the single-commodity case, the arc capacity is
properties of individual atoms, molecules, or bulk matter. stochastic and the system capacity (i.e. the maximum
Conventional reliability theories need to be restudied to flow) is not a fixed number. The system unreliability for
be applied to Nano-Engineering. Research on Nano- (+ 1), the probability that the upper bound of the system
Reliability is extremely important due to the fact that na- capacity equals can be computed in terms of upper boun-
no-structure components account for a high proportion of dary points. An upper boundary point is the maximal
costs, and serve critical roles in newly designed products. system state such that the system fulfills the demand. In
In this paper, Shuen-Lin Jeng et al.[1] introduces the con- his paper Yi-Kuei Lin [3] discusses about multicommodi-
cepts of reliability to nano-technology; and presents the ty limited-flow network (MLFN) in which multicommod-
work on identifying various physical failure mechanisms ity are transmitted through unreliable nodes and arcs.
of nano-structured materials and devices during fabrica- Nevertheless, the system capacity is not suitable to be
tion process and operation. Modeling techniques of de- treated as the maximal sum of the commodity because
gradation, reliability functions and failure rates of nano- each commodity consumes the capacity differently. In
systems have also been discussed in this paper. this paper, Yi-Kuei Lin defines the system capacity as a
Engineers are required to help increase reliabili- demand vector if the system fulfils at most such a de-
ty, while maintaining effective production chedulesto mand vector. The main problem of this paper is to meas-
produce current, and future electronics at the lowest ure the quality level of a MLFN. For this he proposes a
possible cost. Without effective quality control, devices new performance index, the probability that the upper
dependent on nanotechnology will experience high man- bound of the system capacity equals the demand vector
ufacturing costs, including transistors which could result subject to the budget constraint, to evaluate the quality
in a disruption of the continually steady Moores law. level of a MLFN. A branch-and-bound algorithm based
Nano Technology can potentially transform civilization. on minimal cuts is also presented to generate all upper
Realization of this potential needs a fundamental under- boundary points in order to compute the performance
standing of friction at the atomic scale. Furthermore, the index.
tribological considerations of these systems are expected In a computer network there are several reliabili-
to be an integral aspect of the system design and will de- ty problems. The probabilistic events of interest are:
pend on the training of both existing and future scientists, * Terminal-pair connectivity
and engineers in the nano scale. As nanotechnology is * Tree (broadcast) connectivity
gradually being integrated in new product design, it is * Multi-terminal connectivity
important to understand the mechanical and material These reliability problems depend on the net-
properties for the sake of both scientific interest and engi- work topology, distribution of resources, operating envi-
neering usefulness. The development of nanotechnology ronment, and the probability of failures of computing
will lead to the introduction of new products to the pub- nodes and communication links. The computation of the
lic. In the modern large-scale manufacturing era, reliabili- reliability measures for these events requires the enume-
ty issues have to be studied; and results incorporated into ration of all simple paths between the chosen set of nodes.
the design and manufacturing phases of new products. The complexity of these problems, therefore, increases
Measurement and evaluation of reliability of nano- very rapidly with network size and topological connectiv-
devices is an important subject. New technology is devel- ity. The reliability analysis of computer communication
oped to support the achievement of this task. As noted by networks is generally based on Boolean algebra and
Keller, et al. [2], with ongoing miniaturization from probability theory. Raghavendra, et al. [4] discusses vari-
MEMS towards NEMS, there is a need for new reliability ous reliability problems of computer networks including
concepts making use of meso-type (micro to nano) or ful- terminal-pair connectivity, tree connectivity, and multi-
ly nano-mechanical approaches. Experimental verification terminal connectivity. In his paper he also studies the
will be the major method for uvalidating theoretical mod- dynamic computer network reliability by deriving time-
els and simulation tools. Therefore, there is a need for dependent expressions for reliability measures assuming
developing measurement techniques which have capabili- Markov behavior for failures and repairs. This allows
ties of evaluating strain fields with very local (nano-scale) computation of task and mission related measures such as
resolution. mean time to first failure(MTFF) and mean time between
failures (MTF).
2.2 Computer Communication Network A computer communication network (CCN) can
IJSER 2011
205
ISSN 2229-5518
be represented by a set of centers and a set of communica- which failure correlation in common a communication
tion links connecting centers which are up. The network channel is taken into account. In this reliability optimiza-
can be represented mathematically by a graph with nodes tion problem, the assessment on the grid service reliabili-
representing centers and edges representing links. If links ty & performance is a critical component in obtaining the
are simplex, then the graph is directed. Various methods objective function. However, due to the size & complexity
have been derived to evaluate the system reliability and of grid service systems, the existing models for distri-
terminal reliability of a CCN. The methods usually in- buted systems cannot be directly applied. Thus, a virtual
volve the enumeration directly or indirectly of all the tree structure model was developed as the basis of the
events that lead to successful communication between optimization model and a genetic algorithm was adapted
computer centers under consideration. However, little to solve this type of optimization problem for grid task
has been done in optimizing networks except for terminal partition and distribution. In this paper author studied a
reliability. This is because as the number of links increas- case considering different numbers of resources. The Ge-
es, the number of possible assignments to links of the sys- netic Algorithm proved to be effective in accommodating
tem grows faster than exponentially. Kiu Sun-wah et al. various conditions including limited or insufficient re-
[5] present both mathematical and heuristic rules for op- sources.
timizing the system reliability of a CCN with a fixed to- Gregory Levitin et al. [7] in his paper described
pology when a set of reliabilities is given. The techniques grid computing systems with star architectures in which
can be used to predict the system reliability for alternative the resource management system (RMS) divides service
topologies if the network topology is not fixed. The eval- tasks into subtasks and sends the subtasks to different
uation of the terminal reliability of a given computer specialized resources for execution. To provide the de-
communication network is a NP-hard problem. Hence, sired level of service reliability, the resource management
the problem of assigning reliabilities to links of a fixed system (RMS) can assign the same subtasks to several
computer communication network topology to optimize independent resources for parallel execution. Some sub-
the system reliability is also NP-hard. Author develops a tasks cannot be executed until they have received input
heuristic method to assign links to a given topology so data, which can be the result of other subtasks. This im-
that the system reliability of the network is near optimal. poses precedence constraints on the order of subtask ex-
His method provides a way to assign reliability measures ecution. Also the service reliability and performance in-
to the links of a network to increase overall reliability. dices are introduced and a fast numerical algorithm for
their evaluation given any subtask distribution is sug-
2.3 Grid Computing System gested. The sharing that we are concerned with is not
GRID computing is a newly developed technology for primarily file exchange but rather direct access to com-
complex systems with large-scale resource sharing, wide- puters, software, data, and other resources. This is re-
area communication, and multi-institutional collabora- quired by a range of collaborative problem-solving and
tion. This technology attracts much attention. Many ex- resource-brokering strategies emerging in industry,
perts believe that the grid technologies will offer a second science, and engineering. This sharing is controlled by the
chance to fulfill the promises of the internet. The real, Resource Management System (RMS). The Open Grid
specific problem that underlies the Grid concept is coor- Services Architecture enables the integration of services
dinated resource sharing, and problem solving in dynam- and resources across distributed, heterogeneous, dynam-
ic, multi-institutional virtual organizations. Grid technol- ic virtual organizations; and also provides users a plat-
ogy is a newly developed method for large-scale distri- form to easily request grid services. A grid service is de-
buted systems. This technology allows effective distribu- sired to execute a certain task under the control of the
tion of computational tasks among different resources RMS. To provide the desired level of service reliability,
presented in the grid. the RMS can assign the same subtasks to several inde-
Yuan-Shun Dai [6], describe a grid computing pendent resources of the same type. To evaluate the qual-
systems in which the resource management systems ity of service, its reliability and performance indices
(RMS) can divide service tasks into execution blocks (EB) should be defined. Author considers the indices service
and send these blocks to different resources. To provide a reliability (probability that the service task is accom-
desired level of service reliability, the RMS can assign the plished within a specified time), and conditional expected
same EB to several independent resources for parallel system time; and presents the numerical algorithm for
(redundant) execution. According to the optimal schedule their evaluation for arbitrary subtask distribution in a
for service task partition and distribution among re- given grid with a star architecture taking into account
sources, one can achieve the greatest possible expected precedence constraints on the sequence of subtask execu-
service performance (i.e. least execution time), or reliabili- tion. Some of the very helpful practical applications are as
ty. For solving this optimization problem, the author sug-
gests an algorithm that is based on graph theory, Baye- Comparison of different resource management alterna-
sian approach and the evolutionary optimization ap- tives (subtask assignment to different resources),
proach. A virtual tree-structure model is constructed in Making decisions aimed at service performance im-
IJSER 2011
206
ISSN 2229-5518
provement based on comparison of different grid struc- measure that describes the degree of system reliability.
ture alternatives and Distributed program reliability (DPR) is the probability
Estimating the effect of reliability and performance var- that a given program can be run successfully and will be
iation of grid elements on service reliability and perfor- able to access all of the files it requires from remote sites
mance. in spite of faults occurring among the processing ele-
ments & communication links. The second measure, dis-
2.4 Statistical Moments / Bayes Approach tributed system reliability (DSR), is defined as the proba-
In many practical Engineering circumstances, systems bility that all of the programs in the system can be run
reliability analysis is complicated by the fact that the fail- successfully. A distributed system is a collection of pro-
ure time distributions of the constituent subsystems can- cessor-memory pairs connected by communication links.
not be accurately modeled by standard distributions. Ge- The reliability of a distributed system can be expressed
rard L. Reijns et al.[8] in their paper, discuss a low-cost, using the distributed program reliability and distributed
compositional approach based on the use of the first four system reliability analysis. The computing reliability of a
statistical moments to characterize the failure time distri- distributed system is an NP-hard problem. The distribu-
butions of the constituent components, subsystems and tion of programs and data-files can affect the system re-
top-level system. The approach is based on the use of liability. The reliability-oriented task assignment prob-
Pearson Distributions as an intermediate analytical ve- lem, which is NP-hard, is to find a task distribution such
hicle, in terms of which the constituent failure time distri- that the program reliability or system reliability is max-
butions are approximated. The analysis technique is pre- imized. For example, efficient allocation of channels to the
sented for -out-of- systems with identical subsystems, different cells can greatly improve the overall network
series systems with different subsystems and systems throughput, in terms of the number of calls successfully
exploiting standby redundancy. The technique consistent- supported. Chin-Ching Chiu et al.[10] presents a genetic
ly exhibits very good accuracy (on average, much less algorithm-based reliability-oriented task assignment me-
than 1 percent error) at very modest computing cost. In thodology (GAROTA) for computing the DTA reliability
his paper, he present a low-cost, compositional approach problem. The proposed algorithm uses a genetic algo-
based on the use of independent of system size. In addi- rithm to select a program and file assignment set that is
tion, to improve their approach numeric implementation maximal, or nearly maximal with respect to system relia-
details have been outlined by him and a number of ex- bility. Their numerical results show that the proposed
ample applications from the aerospace domain have been algorithm may obtain the exact solution in most cases,
presented in their paper. and the computation time seems to be significantly short-
Pandey et al. [9] provides a Bayes approach of er than that needed for the exhaustive method. The tech-
drawing inference about the reliability of a 1-component nique presented in his paper would be helpful for readers
system whose failure mechanism is simple stress- to understand the correlation between task assignment
strength. The Bayes estimator of system reliability is ob- reliability and distributed system topology. A distributed
tained from data consisting of random samples from the system is defined as a system involving cooperation
stress and strength distributions, assuming each one is among several loosely coupled computers (processing
Weibull. The Bayes estimators of the four unknown shape elements).The system communicates (by links) over a
and scale parameters of stress and strength distributions network. A distributed program is defined as a program
are also considered and these estimators are used in esti- of some distributed system which requires one or more
mating the system reliability. The priors of the parameters files. For a successful distributed program, the local host
of stress and strength distributions are assumed to be possesses the program, the processing elements possess
independent. The Bayes credibility interval of the scale the required files and the interconnecting links must be
and shape parameters is derived using the joint posterior operational.
of the parameters.
2.6 Designed Experiments
2.5 Genetic Algorithm Design of experiments is a useful tool for improving the
Distributed Systems (DS) have become increasingly quality and reliability of products. Designed experiments
popular in recent years. The advent of VLSI technology are widely used in industries for quality improvement. A
and low-cost microprocessors has made distributed com- designed experiment can also be used to efficiently search
puting economically practical. Distributed systems can over a large factor space affecting the products perfor-
provide appreciable advantages including high performance, and identify their optimal settings in order to im-
mance, high reliability, resource sharing and extensibility. prove reliability. Several case studies are available in the
The potential reliability improvement of a distributed literature.
system is possible because of program and data-file re- V. Roshan Joseph et al.[11] presents the devel-
dundancies. To evaluate the reliability of a distributed opment of a integrated methodology for quality and re-
system, including a given distribution of programs and liability improvement when degradation data are availa-
data-files, it is important to obtain a global reliability ble as the response in the experiments. The noise factors
IJSER 2011
207
ISSN 2229-5518
affecting the product are classified into two groups which separable constraints and non decreasing functions. For
led to a Brownian motion model for the degradation cha- illustration, 3 types of problems are solved using his me-
racteristic. A simple optimization procedure for finding thod. A majority of problems concerning system reliabili-
the best control factor setting is developed using an inte- ty optimization are nonlinear programming problems
grated loss function. In general, reliability improvement involving integer variables. The solution methods for
experiments are more difficult to conduct than the quality such problems can be categorized into:
improvement experiments. This is mainly due to the diffi-
culty of obtaining the data. Reliability can be defined as i) Exact methods based on dynamic programming, impli-
quality over time and therefore in reliability improvement cit enumeration and branch-and-bound technique
experiments we need to study the performance of the ii) Approximate methods based on linear and nonlinear
product over time as opposed to just measuring the quali- programming techniques,
ty at a fixed point of time. Two types of data are usually iii) Heuristic methods which yield reasonably good solu-
gathered in reliability experiments: lifetime data and de- tions with little computation
gradation data. Lifetime data gives the information about Each category has both advantages and disadvantages.
the time-to-failure of the product. In the degradation da- Due to the tremendous increase in the available compu-
ta, a degradation characteristic is monitored throughout ting power, the exact solution deserves attention from
the life of the product. Thus, they provide the complete researchers.
history of the products performance in contrast to a sin- To derive an exact solution for a reliability opti-
gle value reported in the lifetime data. Therefore, the de- mization problem, dynamic programming can be used
gradation data contain more information than the lifetime only for some particular structures of the objective func-
data. There are some similarities between reliability im- tion and constraints. It is not useful for reliability optimi-
provement, and quality improvement. Generally speak- zation of a general system, and its utility decreases with
ing, improving the quality will also improve the reliabili- the number of constraints.
ty. But this may not be true always. For example, suppose The Reliability of a Distributed Computing System is the
in a printed circuit board (PCB) manufacturing industry, probability that a distributed program which runs on
tin plating is a more stable process than gold plating. multiple processing elements and needs to communicate
Therefore in terms of improving quality the industry with other processing elements for remote data files will
should prefer tin plating compared to gold plating be- be executed successfully. This reliability varies according
cause a better platted thickness can be achieved using tin to (1) the topology of the distributed computing system,
plating. On the other hand, during customer usage, the (2) the reliability of the communication links, (3) the data
tin will wear out faster than gold and therefore the gold- files and program distribution among processing ele-
plated PCB will have higher reliability. Therefore, gold ments, and (4) the data files required to execute a pro-
plating should be preferred for improving reliability. gram. Thus, the problem of analyzing the reliability of a
Thus, the choice that is good for quality need not always distributed computing system is more complicated than
be good for reliability. Because of this reason, the author the K-terminal reliability problem. In his paper, Lin et al.
should find the procedure for the optimal setting of the [13] describe several reduction methods for computing
factors considering both quality and reliability and also the reliability of distributed computing systems. These
interaction between them. reduction methods can dramatically reduce the size of a
distributed computing system, and therefore speed up
2.7 Distributed and Coherent Systems the reliability computation.
A Design engineer often tries to improve system reliabili- The reliability of a distributed computing system de-
ty with a basic design to the largest extent possible subject pends on the reliability of its communication links and
to several constraints such as cost, weight and volume. nodes and on the distribution of its resources, such as
The system reliability can be improved either by using programs and data files. Many algorithms have been
more reliable components or by providing redundant proposed for computing the reliability of distributed
components. If for each stage of the system, several com- computing systems, but they have been applied mostly to
ponents of different reliabilities and costs are available in distributed computing systems with perfect nodes. How-
the market or redundancy is allowed, then the designer ever, in real problems, nodes as well as links may fail.
faces a decision exercise which can be formulated as a Min-Sheng Lin et al. [14, 15] propose in his paper, two
nonlinear integer programming problem. V. Rajendra new algorithms for computing the reliability of a distri-
Prasad et al. [12] in their paper deals with a search me- buted computing system with imperfect nodes. Algo-
thod based on: rithm I is based on a symbolic approach that includes two
lexicographic order and passes of computation. Algorithm II employs a general
an upper bound on the objective function, factoring technique on both nodes and edges. He also
for solving redundancy allocation problems in shows the Comparisons between both algorithms. It
coherent systems. Such problems generally belong to the shows the usefulness of the proposed algorithms for
class of nonlinear integer programming problems with computing the reliability of large distributed computing
IJSER 2011
208
ISSN 2229-5518
systems. straint is satisfied. Reliability computation is performed

In his paper, Raghavendra, C. S. et al. [16] present a relia- only once, thereby saving much time and the absolute
bility of a distributed processing system is an important error of the proposed algorithm from exact solution is
design parameter that can be described in terms of the very small.
reliability of processing elements and communication
links and also of the redundancy of programs and data
2.8 Heuristic and General Reduction Methods
files. The traditional terminal-pair reliability does not cap-
ture the redundancy of programs and files in a distri- Distributed Computing Systems (DCS) have become a
buted system. Two reliability measures are introduced: major trend in computer system design today, because of
distributed program reliability, which describes the prob- their high speed and reliable performance. Reliability is
ability of successful execution of a program requiring an important performance parameter in DCS design. In
cooperation of several computers, and distributed system the reliability analysis of a DCS, the term of K-Node Re-
reliability, which is the probability that all the specified liability (KNR) is defined as the probability that all nodes
distributed programs for the system are operational. in K (a subset of all processing elements) are connected. In
These two reliability measures can be extended to incor- his paper, Ruey-Shun Chen et al. [19, 20] proposed a sim-
porate the effects of user sites on reliability. An efficient ple, easily programmed heuristic method for obtaining
approach based on graph traversal is developed to eva- the optimal design of a DCS in terms of maximizing relia-
luate the proposed reliability measures. bility subject to a capacity constraint. The first half of his
In his paper Deng-Jyi Chen et al. [17] Presents an al- paper presents a heuristic algorithm which selects an op-
gorithm for computing the reliability of distributed com- timal set of K-nodes that maximizes the KNR in a DCS
puting systems (DCS). The algorithm, called the Fast Re- subject to the capacity constraint. The second half of the
liability Evaluation Algorithm, is based on the factoring paper describes a new approach that uses a K-tree disjoint
theorem employing several reliability preserving reduc- reduction method to speed up the KNR evaluation. On
tion techniques. The effect of file distributions, program comparing with existing algorithms on various DCS to-
distributions, and various topologies on reliability of the pologies, he found that the proposed algorithm is a sub-
DCS is studied in brief using the Fast Reliability Evalua- optimal design much more efficiently in terms of both
tion algorithm. Compared with existing algorithms on execution time and space than an exact and exhaustive
various network topologies, file distributions, and pro- method for a large DCS.
gram distributions, the proposed algorithm in his paper is In his paper, Ruey-Shun Chen et al. [21] present a
much more economical in both time and space. simple, easily programmed exact method for obtaining
In his paper, Chiu, Chin Ching et al. [18] describe the optimal design of a distributed computing system in
about Distributed system that provides a cost-effective terms of maximizing reliability subject to memory capaci-
means of enhancing a computer system's performance in ty constraints. he assume that a given amount of re-
areas such as throughput, fault-tolerance, and reliability sources are available for linking the distributed compu-
optimization. Consequently, the reliability optimization ting system. The method is based on the partial order re-
of a distributed system has become a critical issue. A K- lation. To speed up the procedure, some rules are pro-
terminal reliability is defined as the probability that a posed to indicate conditions under which certain vectors
specified set, K, of nodes is connected in a distributed in the numerical ordering that do not satisfy the capacity
system. A K-terminal reliability optimization with an or- constraints can be skipped over. Simulation results show
der (the number of nodes in K-terminal) constraint prob- that the proposed algorithm requires less time and space
lem is to select a K-terminal of nodes in a distributed sys- than exhaustive method.
tem such that the K-terminal reliability is maximal and
possesses sufficient order. It is evident that this is an NP- 2.9 Tele-Communication / Neural Network
hard problem. This paper presents a heuristic method to
reduce the computational time and the absolute error In Studies on the design of communications networks,
from the exact solution. The method proposed is based on reliability has been defined in a number of ways. Fulya
not only a simple method to compute each nodes weight Altiparmak et al. [19], discusses about a probabilistic
and each links weight, but also an effective objective measure i.e. all-terminal reliability, is considered (this is
function to evaluate the weight of node sets. Before ap- sometimes termed overall network reliability). All-
pending one node to a current selected set, instead of terminal reliability is the probability that a set of opera-
computing the weight of all links and all nodes of each tional edges provides communication paths between
set, only the weight of a node, which is adjacent to the every pair of nodes. A communications network is typi-
current selected set, and links between the node and the cally modeled as a graph with N-nodes, and L- edges;
current selected set are accumulated. Then the proposed nodes represent sites (computers), and edges represent
algorithm depends on the maximum weight to find an communication links. Each node, and each edge has an
adequate node and assign it to the current selected set in associated probability of failure, and the reliability of the
a sequential manner until the order of K-terminal con- network is the probability that the network is operational.
IJSER 2011
209
ISSN 2229-5518
In his paper, Fulya Altiparmak et al. [19] propose a new ferent topic such as Distributed and Coherent Systems,
method, based on an artificial neural network (ANN), to Network Reliability Optimization, Tele-
estimate the reliability of networks with identical link Communication / Neural Network etc.
reliability. There are two significant advantages to this.
The first is that a single ANN model can be used for mul-
4 PROPOSED FUTURE WORK
tiple network sizes and topologies. The second advantage
is that the input information to the ANN is compact, The topic reliability remains an interesting challenge for
which makes the method tractable, even for large sized future research. The possible direction is to investigate
networks. We use the approach for networks of widely whether reliability ranking for improving the perfor-
varying reliability, and then consider only highly reliable mance of the system algorithm. Also to find the better
networks. reliability system in the sense of Distributed computing
System/ Network system it is necessary to improve the
2.10 Network Reliability Optimization time complexity for such system. In the sence of nano-
Evaluation of the reliability of a network is a fundamental technology much work is needed in particular nano-
problem. It has application in many practical fields such reliability field to ensure the product reliability and safety
as communication, digital systems, and transportation is various use condition. Some other meta-heuristic ap-
systems. The physical network is represented by a graph proaches besides the above may also be applicable such
composed of nodes connected by directed and undirected as Tabu-Search, Hybrid Optimization Technique and Ant
arcs. Associated with each arc and with each node of a Colony Optimization etc. Future aspects also exist in var-
graph is a failure probability. Debany, W. H. et al. [23] ious communication schemes in Wireless (COBRA:
provide a graph with known failure probabilities of its Common Object Request Broker Architecture). This is
elements (arcs and nodes) the objective is to find the also easily extensible to generic wireless network system.
probability that at least one complete simple path (no From the point of view of quality management, treat the
node is visited more than once) exists between a source system reliability as a performance index, and conduct
and a terminal node. the sensitive analysis to improve the most important
Developments and Improvements in information component (e.g. transmission line, switch or server) will
& communication technologies in recent years have re- increase the system reliability most significantly. Future
sulted in increased capacities, and higher concentration of research can extend the problem from the single com-
traffic in telecommunication networks. Operating failures modity case to the multicommodity case. Besides, trans-
in such high capacity networks can affect the quality of mission time reduction is a very important issue for an
service of a large number of consumers. Consequently, information system. Therefore researcher can extend the
the careful planning of a networks infrastructure and the work to include the time attribute for each component.
detailed analysis of its reliability become increasingly Reliability can also be extended for hybrid Fault-Tolerant
important toward ensuring that consumers obtain the embedded system architecture in the form of hybrid re-
best service possible. One of the most basic, useful ap- covery Block (RB). Future research will also improved the
proaches to network reliability analysis is to represent the automated controller abilities, and the human machine
network as an undirected graph with unreliable links. interface, in order to increase the efficiency of the human
The reliability of the network is usually defined as the reasoning assistance, and to decrease the human response
probability that certain nodes in the graph are connected time.
by functioning links. Dirk et al. [24] is discusses this
Network reliability Optimization with network planning, ACKNOWLEDGMENT
where the objective is to maximize the networks reliabili- The author wish to thank Prof. P. N. Tondon for the
ty, subject to a fixed budget. They can develop a number support and the encouragement which he yields to
of simulation techniques to address the network reliabili- them during their research work.
ty estimation problem.
REFERENCES
3 CONCLUSION [1] Shuen-Lin Jeng, Jye-Chyi Lu, and Kaibo Wang, A Review of
The behaviour of reliability is much more sensitive to Reliability Research on Nanotechnology, IEEE Transactions on
change in different fields of engineering and sciences. Reliability, Vol. 56, No. 3, Pg. 401-410, September 2007.
Around twenty five papers from leading journals in dif- [2] J. Keller, A. Gollhardt, D. Vogel, and B. Michel, Nanoscale
ferent fields of reliability have been covered in this paper. Deformation Measurements for Reliability Analysis of Sen-
This paper is reviewed various aspects of reliability re- sors, presented at the Proceedings of the SPIEThe Interna-
search in the field of Nano-Technology, Computer tional Society for Optical Engineering, 2005.
Communication Network, Grid Computing System, [3] Yi-Kuei Lin, System Reliability of a Limited-Flow Network in
Statistical Moments / Bayes Approach, Genetic Algo- Multicommodity Case, IEEE Transactions on Reliability, Vol.
rithm etc. We have broken down our survey into the dif- 56, No. 1, Pg. 17-25, March 2007.
IJSER 2011
210
ISSN 2229-5518
[4] Raghavendra, C. S., Kumar, V. K. P. and Hariri S., Reliability [20] Ruey-Shun Chen, Deng-Jyi Chen and Y. S. Yeh, Reliability
analysis in Distributed systems, IEEE Transactions on Com- Optimization on the Design of Distributed Computing
puters, Volume 37, Issue 3, Pg. 352 358, March 1988. Systems, Proceed of Int. Conf. on Computing and Information,
[5] Kin-Sun-Wah and McAlister D.F, Reliability optimization of Pages 422-437, 1994.
computer communication network, IEEE Trans. On Reliability, [21] Ruey-Shun Chen, Deng-Jyi Chen and Y. S. Yeh, Reliability
Vol.37, No. 2, Pp.275- 287 (1998) Dccember. Optimization of Distributed Computing Systems Subject to
[6] Yuan-Shun Dai, Optimal Resource Allocation for Maximizing Capacity Constraints, Computers & Mathematics with
Performance and Reliability in Tree-Structured Grid Services, Applications, Volume 29, Issue 4, Pages 93-99,February 1995.
IEEE Transactions on Reliability, Vol. 56, No. 3, Pg. 444-453, [22] Fulya Altiparmak, Berna Dengiz, and Alice E. Smith, A General
September 2007. Neural Network Model for Estimating Telecommunications
[7] Gregory Levitin, Yuan-Shun Dai, and Hanoch Ben-Haim, Re- Network Reliability, IEEE Transactions on Reliability, Vol. 58,
liability and Performance of Star Topology Grid Service With No. 1, Pg. 2-9, March 2009.
Precedence Constraints on Subtask Execution, IEEE Transac- [23] Debany, W. H. and Varshney, P. K., Network Reliability Evalu-
tions on Reliability, Vol. 55, No. 3, Pg. 507-515, September 2006. ation Using Probability Expressions, IEEE Transactions On Re-
[8] Gerard L. Reijns and Arjan J. C. Van Gemund, Reliability liability, Vol. R-35, No. 2, Pg. 161-166, 1986 June.
Analysis of Hierarchical Systems Using Statistical Moments, [24] Dirk P. Kroese, Kin-Ping Hui, and Sho Nariai, Network Relia-
IEEE Transactions On Reliability, Vol. 56, No. 3, Pg 525-533, bility Optimization via the Cross-Entropy Method, IEEE
September 2007. Transactions on Reliability, Vol. 56, No. 2, Pg. 275-287, June
[9] Pandey, M. and Upadhayay S. K., Reliability Estimation in
2007.
Stress- Strength Models: A Bayes Approach, IEEE Transac-
tions on Reliability, December 1985.
[10] Chin-Ching Chiu, Chung-Hsien Hsu, and Yi-Shiung Yeh, A
About the Author: Dr. Anju Khandelwal is an as-
Genetic Algorithm for Reliability-Oriented Task Assignment
sistant professor and HOD-Dean Academics of
With k Duplications in Distributed Systems, IEEE Transactions
S.R.M.S. Womens College of Engg. & Tech. Bareilly
On Reliability, Vol. 55, No. 1, Pg. 105-117, March 2006.
(U.P.) Affiliated by G. B. T. U. Lucknow-India. She
[11] V. Roshan Joseph and I-Tang Yu, Reliability Improvement
received her Bachelor degrees in Mathematics and
Experiments With Degradation Data, IEEE Transactions On
Physics from Bundelkhand University Jhansi (U.P.),
Reliability, Vol. 55, No. 1, Pg. 149-157, March 2006.
India in 1996, and Master degree in Mathematics with
[12] V. Rajendra Prasad and Way Kuo, Reliability Optimization of
Computer Application from Bundelkhand University
Coherent Systems, IEEE Transactions On Reliability, Vol. 49,
Jhansi (U.P.), India in 1998. She then joined the Bun-
No. 3, Pg. 323-330, September 2000.
delkhand University Jhansi as a lecturer under Self
[13] M.-S. Lin and D.-J. Chen, General Reduction Methods for the
finance scheme. She completed his PhD degree in
Reliability Analysis of Distributed Computing Systems, The
Operations Research from Gurukula kangri Universi-
Computer Journal, Volume 36, Issue 7, Pages 631-644, 1993.
ty Hardwar (Uttaranchal) in 2006. She received Mas-
[14] Min-Sheng Lin, Deng-Jyi Chen and Maw-Sheng Horng, The
ter degrees in technical field that is M. Tech (Software
Reliability Analysis of Distributed Computing Systems with
Imperfect Nodes, The Computer Journal, Volume 42, Issue 2,
Engineering) in 2010 from U. P. T. U. Lucknow (U.P.),
Pages 129-141, 1999. Her areas of interest include parallel and distributed
systems, optimization techniques, CCN and reliabili-
[15] MIN-SHENG LIN AND DENG-JYI CHEN, GENERAL REDUCTION
ty Analysis. She is a life member of the IAPS.
METHODS FOR THE RELIABILITY ANALYSIS OF DISTRIBUTED
COMPUTING SYSTEMS, COMPUTER J OURNAL ISSN 00104620
CODEN CMPJA6, VOL. 36, NO. 7, PP. 631-644, 1993.
[16] Raghavendra, C. S. and Makam, S. V., Reliability Modeling
and Analysis of Computer Networks, IEEE Transactions on
Reliability, Vol. R-35, No. 2, Pg. 156-160, 1986 June.
[17] Deng-Jyi Chen and Min-Sheng Lin, On Distributed Compu-
ting Systems Reliability Analysis Under Program Execution
Constraints, IEEE Transactions on Computers, Volume 43, Is-
sue 1, Pages 87-97, 1994.
[18] Chiu, Chin Ching , Yeh, Yi Shiung and Chou, Jue Sam, An
Effective Algorithm for Optimal K-terminal Reliability of Dis-
tributed Systems, Malaysian Journal of Library & Information
Science, 6 (2), Pg. 101-118, 2001.
[19] Ruey-Shun Chen, Deng-Jyi Chen and Y. S. Yeh, A New Heuris-
tic Approach for Reliability Optimization of Distributed Com-
puting Systems subject to Capacity Constraints, Journal of
Computers & Mathematics with Applications, Volume 29, Issue
3, Pages 37-47, Feb 1995.
IJSER 2011
211
ISSN 2229-5518
An Innovative Quality of Service (QOS)

based Service Selection for Service
Orchrestration in SOA
S.Neelavathi and K.Vivekanandan
Abstract Service Oriented Architecture (SOA) has become a new software development paradigm because it provides a
flexible framework that can help reduce development cost and time. SOA promises loosely coupled interoperable and
composable services. Service selection in business processes is the usage of techniques in selecting and providing quality of
services (QOS) to consumers in a dynamic environment. Single business process model consists of multiple service invocations
forming service orchestration. It represents multiple execution paths called modeled flexibility. In certain cases, modeled
flexibility can cause conflicts in service selection optimization, making it impossible to simultaneously optimize all execution
paths. This paper presents an innovative approach to service selection for service orchestration that addresses this type of
conflicts by encompassing status identification based availability estimation with multiple QOS constraints along with an
effective quality assessment model. This model captures the expectations from the users on the multiple quality of a service and
returns ratings as a feedback on the service usage. This updated rating in the service list can be used by the new user. This
proposed method provides optimal services to users consistently and efficiently thereby resulting in more meaningful and
reliable selection of services for service orchestration in SOA.
Index Terms Service Oriented Architecture, Service Selection, Service orchestration, Meta-metrics, Modeled Flexibility,
Rating, Multiple QOS level, local selection, global selection.
1 INTRODUCTION
Service oriented architecture (SOA) is a new para- Status identification based availability estimation for
digm for software development that promises loosely service selection is used along with multiple QoS con-
coupled, interoperable and composable components straints. It is extended with an effective quality assess-
called services. Service orchestration is the execution of a ment model that is used to match the expectations from
single transaction that impacts one or more services in an the user with that of rating of services held in service list.
organization. It is called as business process. Business The service list is divided into four groups with all the
processes are implemented by orchestrating services of services having the triple factor of quality rating for all
different activities involved in it. Multiple QoS-based ser- the multiple QoS constraints. The feedback from the user
vice selection results in selecting an optimal service for after service usage is used to update the service list. This
single activity from a set of candidate services, thereby proposed approach of service selection with multiple QoS
maximizing the QoS of the entire business process. A sin- factors results in more meaningful and reliable selection
gle business process model can represent multiple execu- of services used in service orchestration in service
tion paths known as modeled flexibility. Modelled flex- oriented architecture. Along with this, it also resolves the
ibility can cause conflicts in service selection if the differ- conflicts with modeled flexibility in business process by
ent optimal services are selected for the common activity meta-metrics thereby ensuring selection of optimal ser-
in both the execution paths thereby making it impossible vices for all the service invocations.
to optimize all execution paths. The proposed approach
to service selection addresses this type of conflicts
through a set of meta-metrics (probability of execution).
2 SERVICE ORIENTED ARCHITECTURE
The service orientation and Service Oriented Architecture
(SOA) are not new or revolutionary concepts. It is the
K.Vivekanandan is Professor in the Department of Computer Science and next stage of evolution in the distributed computing [2].
Engineering at Pondicherry Engineering College, Puducherry.Mobile:
9443777795 Fax: 2655101 Mail id: [email protected].
SOA is not a Technology and is only an architectural ap-
S.Neelavathi is Assistant Professor in the department of Information Tech- proach. SOA is defined as Service Oriented Architecture
nology, PKIET, Karaikal. Mobile: 9443632464. Mail Id: neela- (SOA) is a paradigm for organizing and utilizing distri-
[email protected] buted capabilities that may be under the control of differ-
ent ownership domains [2]. SOA includes the previously
proven and successful elements from past distributed
IJSER 2011
212
ISSN 2229-5518
paradigms. These elements are combined with the design tions. The agnostic nature of services enables them to
approaches to leverage recent technology in distributed be recombined and reused in different forms.
computing [1]. Service Autonomy
In SOA, the loosely coupled systems do compu-
ting in terms of services. SOA separates functions into Services to carry out their functionalities consistently
distinct units, or services, which developers make access- and reliably, needs to have a significant degree of
ible over a network in order to allow users to combine control over its environment and resources.
and reuse them in the production of applications. These Service Statelessness
services communicate with each other by passing data
from one service to another [3], or by coordinating an Services are designed to remain stateful only when
activity between two or more services. They use the well required. If there are stateful, then the management
established standards [4]. of excessive state information can compromise the
This approach is based on the design principles of availability of the service.
loose coupling, which is a principle by which the con- Service Discoverability
sumer and service are insulated from changes in underly-
ing technology and behavior, Interoperability[2] the prin-
Service discovery is the process of discovering a
ciple which provides the ability to support consumers
service and interpretation is the process of under-
and service providers that are of different programming
standing its purpose and capabilities.
languages on, different operating systems with different
Service Composability
communication capabilities, Encapsulation that allows the
potential consumer to be insulated from the internal
technology and even the details of behavior of service, It is related to its modular structure. It is com-
Discoverability which is used to realize the benefit of posed in three ways. Application is an assembly of
reuse [5]. Seamless integration of various systems allows services, components and application logic that binds
data access from anywhere anytime, thereby providing functions together. Service Federation is the collection
services to customers and partners inside and outside the of services managed together in a large service do-
enterprise. It provides a simple scalable paradigm for main and Service orchestration is execution of single
organizing large networks of systems that require intero- transaction that impacts one or more services in an
perability and develops systems that are scalable, evolva- organization.
ble and manageable and establishes solid 2.2 Elements of SOA
foundation for business agility and adaptability.
The overall architectural model in the Fig 1
shows the elements of SOA. A service provider describes
2.1 Services In SOA
its service using Web Service Description Language
The most fundamental unit of service oriented (WSDL). The WSDL definition is divided into two parts:
solution logic is the service. Services in SOA comprises the abstract description that defines the service interface,
the below 8 distinct design principles. and the concrete description that establishes the transport
Standardized service contract and location information. This WSDL definition is pub-
lished to the Universal Discovery Description Interface
Services express their purpose and capabilities via a (UDDI) service registry. SOAP (Simple Object Access Pro-
service contact. It is the most fundamental part of service- tocol) is the universally accepted standard transport pro-
orientation. tocol for messages and represents a standardized format
Service Loose Coupling for transporting messages. SOAP message contents are
presented in message body which consists of XML for-
Coupling refers to the number of dependencies be- matted data. A service requestor issues one or more que-
tween modules. Loosely coupled modules have a few ries to the UDDI to locate a service and determine how to
known dependencies whereas tightly coupled have many communicate with that service. Part of the WSDL pro-
unknown dependencies. SOA promotes loose coupling vided by the service provider is passed to the service re-
between service consumers and providers. questor. This tells the service consumer what the requests
Service Abstraction and responses are for the service provider. The service
consumer uses the WSDL to send a request to the service
This emphasizes the need to hide as much of the un- provider. The service provider provides the expected re-
derlying details of a service as possible to preserve sponse to the service consumer.
loosely coupled relationship.
Service Reusability
The same services can be reused in multiple applica-

IJSER 2011
213
ISSN 2229-5518
3.1 Quality of service
Quality of service is a combination of several

Service
qualities or properties of a service such as Performance,
Registry
Cost, Reliability, Availability, Reputation, and Fidelity.
The performance is the time duration from a request be-
ing sent, to when the results are received. Cost refers to
the amount of money that the consumer pays for using a
Discover and re- service. The reliability is the probability that the re-
SOAP Publish quested service is working without a failure within a spe-
trieve WSDL
mes- cified time frame. Availability is the quality aspect of
WSDL
sages whether the service is present or ready for immediate use.
Reputation is the criterion in measuring total trustworthi-
ness of a service and Fidelity is the average marks that are
given by different consumers to the same QoS criterion.
Long term relationship between service provider
Service Request based on and service consumer generate Service Level Agreement.
Service
requestor WSDL The service consumer may impose several constraints to
Provider
be satisfied for the services utilized thereby generating
SLA. For Example the cost of service is less than 200$,
response time is less than 2ms and service should be
Response based on available less than 90% of the time etc.
WSDL
3.1 Classificatiion of service selection methods
Service selection is classified as static service se-
Figure 1 Elements of SOA
lection and dynamic service selection. The further is selec-
tion of optimal services from a set of services with similar
3 SERVICE SELECTION IN SOA functionalities during runtime while the later is selection
before the runtime. Another scenario is to select services
It is possible to have multiple service providers in an
among several group of services, with each group having
SOC environment who offer broadly similar services but
similar services called as Local selection and Global selec-
with differing qualities. It is essential, therefore, that the
tion. Local Selection is to compose a system that can meet
selection process should select optimal service out of a
certain QoS criteria with selecting one optimal service
group of functionality-similar services optimized for a
from each group also referred as service composition
certain property of QoS. The aim of service selection is to
whereas Global Selection does not aim at selecting an op-
provide meaningful and reliable services to consumers
timal service in each group, but optimizing the QoS of the
consistently and efficiently from the set of candidate ser-
composed system. Table I shows the references of static
vices. The figure 2 depicts a typical scenario of service
service selection and Table II shows the classification of
selection. When the selector receives a request from a
dynamic service selection methods.
client, the selector chooses a service to invoke and returns
the response message to the client. The selection process
TABLE I
can be optimized for a certain property of QoS.
REFERENCES OF STATIC SERVICE SELECTION
Pattern recognition based adaptive catego-

rization technique and solution for services
static ser- selection IEEE2007.
vice selec- A Petri-net based specification model for
tion Web services In Proceedings of ICWS
2004,IEEE
Automatic Web service composition based
on graph network analysis metrics Internet
Systems 2005 Springer, 2005.
TABLE II
Figure 2 A Typical Scenario of Service Selection REFERENCES OF DYNAMIC SERVICE SELECTION
IJSER 2011
214
ISSN 2229-5518
TABLE II
REFERENCES OF DYNAMIC SERVICE SELECTION
Roland Ukor, Andy Carpenter , Flexible Service Selection Optimization Using Meta-metrics Congress on
Based on QoS Services-I, IEEE 2009
Under no global V.Deora,j.Shao,w.A.Gray, supporting qos based selection in service oriented architecture proceedings of the
constraint international conference on next generation web services practices , IEEE 2006
This selection
process selects Y wang, J Yang,Relation Based Service Networks for reliable service selection proceedings of the conference
optimal service out on commerce and enterprise computing,IEEE 2009
of a group of func- Canfora, G., Di Penta, M., Esposito, R., and Villani, M. L.,An Approach for QoS-aware Service Composi-
tionality-similar tion based on Genetic Algorithms, Proc. of the 2005 Conf. on Genetic andevolutionary computation,
services optimized ACM Press, New York, 2005.
for a certain prop- Under single global
erty of QoS. constraint D.Liu, Z.Shao,C.Yu, A heuristic Qos-aware service selection approach to web service composition, Interna-
tional Conference on computer and Information science IEEE 2009
Lingshuang S,Lu Z, et al. Dynamic Availability Estimation For Service Selection Based On Status Identifi-
cation, IEEE International Conference on web Services, 2008 IEEE
Bang y, chi-Hung,et al. Service selection model based on QoS reference vector, Congress on services , IEEE
2007
D.A.Menasce et al. On optimal service selection in SOA Performance Evaluation 67 (2010) 659-675
Under multiple V.Diamadopoulau et al.Techniques to support Web Service selection Journal of Network and Computer
global constraint Applications (2008)
D.Liu, Z.Shao,C.Yu, A heuristic Qos-aware service selection approach to web service composition, Interna-
tional Conference on computer and Information science IEEE 2009
Based on Semantic Achieves the similari- Z Guoping, Z Huijuan, Wang Z An Approach to QoS-aware service selection in Dynamic Web service composi-
web ty comparison by tion IEEE(2007)
calculating the se-
mantic distance by V.X.Tran et al. QoS ontology and its QoS-based ranking algorithm for Web services Simulation Modelling
QOS and context. Practice and Theory 17 (2009)
Based on improving Add new actions to Balke W.T, Wagner M, Kim S.M, et (2004)
standard UDDI to
protocol or lan-
achieve dynamic
guage UDDI process, or B.Jeong et al. On the functional quality of service (FQoS) to discover and compose interoperable web services
design a selecting Expert Systems with Applications (2009)
language like SQL,
select Web services
by setting restrictive
conditions.
Based on user pre- Through users' scores O.Minhyuk, B.Jongmoon,et.al, An efficient approach for QoS-aware service selection based on a tree-based
on web services,
ference algorithm Seventh IEEE international conference on computer and Information science 2008
achieves dynamic
updating of Web
services selection TQOS for automatic web service composition IEEE transactions on services computing (2010)
system, thus forming
a dynamic selection
process with self-
evaluation function.
IJSER 2011
215
ISSN 2229-5518
4 PROPOSED QOS BASED SERVICE possible to obtain a solution that is simultaneously op-
SELECTION FOR SERVICE timal for both paths, resulting in problems with service
ORCHESTRATION selection. This raises conflicts with modeled flexibility.
In order, to resolve these conflicts with multiple
Service orchestration is the execution of a single execution paths in business process, a set of process me-
transaction that impacts one or more services in an organ- trics called meta-metrics is used. They exist solely for the
ization, called a business process. In order to maximize purpose of biasing the evaluation of QoS metrics for can-
the benefits of SOA, service selection is important espe- didate services based on priorities. With probability of
cially in terms of providing quality of services (QoS) to execution as meta-metric, activities with a higher proba-
consumers in a dynamic environment. The proposed bility of being executed are given priority to resolve the
work is intended to develop a more effective, meaningful selection conflicts between the optimal solutions for two
and robust service selection methodology in service or- or more execution paths.
chestration for conflict resolving using Meta metrics .It
aims at selecting reliable and optimal services by using 4.2 STATUS IDENTIFICATION BASED AVAILABILITY
more than one relevant QoS category along with an effec-
tive quality assessment model. ESTIMATION FOR SERVICE SELECTION (SIBE)
4.1 Business Process The architecture of SIBE is shown in Figure 4. Quality

of service assessment model is used to capture the rating
Business Process consists of multiple activities,
from the user for the service invoked. The rating is a triple
each of which is a service invocation. If any of the service
factor consisting of the expectation from the user for all
invocation fails, the entire transaction should be rolled
the five QOS criterias namely cost, performance, reliabili-
back to the state that existed before execution of the
ty, reputation and fidelity ,along with the provision of
transaction. Let P be a business process with a set of activ-
actual perceived rating and quality rating offered by the
ities A= {a1, a2,.an}. For each activity ai A, there exists
user after using the service. When the user request is
a set of candidate services Si = {s1, s2,.sn} where any of
made for a service in the form of triple factor, the selector
the optimal service can be used for the activity. A candi-
tries to pick up an optimal service from the service list.
date service is selected based on multiple QoS metrics
such as cost, performance, reliability, reputation and fidel- The selector maintains four lists of services. Each list
represents a separate status stable up, transient down,
ity. A solution to the service selection problem is
short term down and long term down. First the selector
represented as a pair {(ai,si)} which assigns a service si Si
searches for the optimal service in the stable up list. If the
for each activity ai. The Figure 3 shows a business process
stable up list is empty, the selector tries the transient
with four activities. The optimal service is selected for the
down list. Finally the selector searches the short term
business dynamically on an instance-by-instance basis.
down list and still furthers the long term list. Whenever
one service status has been identified, the service is in-
serted into the appropriate list. The service in the service
list is presented along with expectation, actual perceived
value and feedback rating for all the QoS attributes from
the set of users. Availability which indicates the probabili-
ty of successful invocation of a service is a key property
used in service selection. Multiple QOS criteria such as
cost, performance, reliability, reputation and fidelity of
the service are taken into consideration.
When the selection process ends, the selector puts the
invocation records into historical records and reevaluates
the status of the invoked services. The fragmentor algo-
rithm takes invocation records for each service from the
historical records and fragments it into equal number of
fragments by randomly chosen fragment length. Each
fragmented segment consists of results of service invoca-
tions in terms of either success or failure. The categorizer
Figure 3 Example Business process uses exponentially weighted average to categorize the
services into appropriate list in the service list which is in
A single business process model can represent turn used by the selector to select the optimal service
multiple execution paths, a phenomenon called modeled based on user expectation.
flexibility. An optimal solution for path1 may select can- The user provides a feedback on the service usage in
didate s12 for a1, while the optimal solution for path2 se- terms of quality rating for all the QoS criterias. The ser-
lects a different candidate s11 for a1. Therefore, it is im- vice list is updated with the quality rating offered by the
IJSER 2011
216
ISSN 2229-5518
recent user. The new service user receives changes from attributes, thereby providing a more meaningful and reli-
the service list and uses this updated rating for reselecting able service as the output. The actual rating value i.e
their optimal services. As the inclusion of rating of many feedback provided by the user after the service usage is
new users will subsequently incur more memory space, added into the service list. This updated rating can be
provision is made to remove the older entry thereby pro- used by the user with the same intentions. Further to
viding an efficient method of data storage. maintain a fixed set of entries in the service list, the older
entries are always removed paving way for new entries of
ratings.
Figure 4 Architecture of SIBE
The figure 5 shows the quality of service assess- Figure 5 Quality of service assessment model used in
ment model used in SIBE. It captures the rating from the SIBE
users. The user apart from mentioning the functionalities
of the required service is also prompted to specify quality 4.3 PROPOSED ARCHITECTURE FOR SERVICE SELEC-
rating for all the QoS criterias of the intended service. TION METHOD
Availability is one of the key QoS attribute. Other QoS
attributes considered are cost, performance, reliability,
reputation and fidelity. Users specify quality rating in the
form of triple factor comprising Expectation from the ser-
vice, perceived value from the user and the actual rating
offered by the user after using the service. The triple fac-
tor for cost is denoted as E(c ),P(c ) and R(c ).similarly it is
E(per)P(per)R(per) for performance ,E(r )P(r )R(r ) for re-
liability, E(rep)P(rep)R(rep) for reputation and
E(F)P(F)R(F) for fidelity.
The service held in four lists consists of triple
factor for all the QoS criterias comprising expectation,
perceived value and the actual rating. Among the expec-
tation rating provided by the user for all the QoS
attributes, the highest rating of a particular attribute value Figure 6 Proposed architecture model
is taken. Then the particular attribute with highest rating
is mapped against the attribute of the service in the ser- The above Figure 6 shows the proposed architec-
vice list with the same expectation. If the expectation ture model for service selection in service orchestration of
matches, then the appropriate service from the service list a business process model. The client in the above figure is
is selected and offered to the user. Meeting the expecta- anonymous with the business process. Every business
tion for a single attribute does not mean to satisfy the oth- process model represents multiple execution paths, a
er attributes to a significant extend. So, the expectation condition known as modeled flexibility. Activity analyzer
matching can be extended to a maximum number of other analyzes each execution paths for the total number of ac-
IJSER 2011
217
ISSN 2229-5518
tivities. For all the activities analyzed, service selection 4. Rating Vs three QoS levels namely low, mod-
optimization is carried out dynamically on an instance- erate and high.
by-instance basis by SIBE(Status Identification based
Availability Estimation for Service Selection)as explained
5. Selection and invocation of services Vs proxy
above.
Each execution path represents a set of activities servers stored.
in a service oriented business process. Each activity in the
business process involves service invocation resulting in
service orchestration for the entire transaction. Service 6.CONCLUSION
selection in a transaction can be carried out once for all
activities of a process, or may be carried out on an activi-
ty-by- activity basis. Here service selection on the latter This paper presents an innovative approach to multiple
grounds is used. The optimal service selected for each Quality of service (QoS) based service selection
activity in all execution paths is held in the candidate ser- SIBE(STATUS IDENTIFICATION BASED AVAILABILITY
vice list. Service orchestration is finally derived by service ESTIMATION FOR SERVICE SELECTION ) for service
aggregations of services stored in candidate service list. orchestration in service oriented architecture. SIBE results
Meta metrics assigner check the solution set in in more meaningful and reliable selection of services used
the candidate service list comprising of activities along in service orchestration. Execution of single transaction
with their optimal services. If path1 and path 2 of busi- that impacts one or more services in an organization is
ness process selects same optimal service for the common called a business process. Modelled flexibility in business
activity, then there is no problem. If there is a mismatch processes can cause conflicts between optimal service
between the services in the execution paths for the com- selections for activities that are common to multiple ex-
mon activity, then there raises conflict of modeled flexibil- ecution paths. This approach presented in this paper ad-
ity. Here priority of execution of activities in execution dresses these conflicts by using a set of meta-metrics
paths is used as meta-metrics to resolve the conflicts. along with meaningful service selection.
Comparison of priority among the activities in different
execution paths is carried out by analyzing the actual REFERENCES
working environment. Then the activity with a higher [1] Thomas Erl, (2005) Service-Oriented Architecture: Concepts,
probability of execution is assigned the optimal service Technology, and Design. Prentice Hall, PTR
and the same optimal service is also allotted to other [2] Judith Hurwitz, Robin Bloor, Baroudi C,Marcia K Service
common activities in different execution paths by the ser- Oriented Architecture for Dummies Wiley Publishing, Inc.
vice adjustor thereby overcoming the conflicts with mod- [3] Soa Principles Of Service Design
eled flexibility. Thus the proposed work represents an [4] Canfora, G., Di Penta, M., Esposito, R., and Villani, M L.,An
innovative approach of service selection for service or- Approach for QoS-aware Service Composition based on Genetic Al
chestration by addressing the conflicts with modeled flex- gorithms, Proc. of the 2005 Conf. on Genetic and evolutionary
ibility based on meta-metrics. computation, ACM Press, New York, 2005.
[5] D. T. Tsesmetzis, I. G. Roussaki, I. V. Papaioannou, et al.QoS
awareness support in Web-Service semantics, in Proceedings of
5.0 EVALUATION METHOD AICT/ICIW 2006, 2006.
Evaluation Environment: [6] Yu and K. Lin, Service selection algorithms for Web services with
1. Test Bed end-to-end QoS constraints, in Proceedings CEC04, pp. 129-
136, 2004.
2. Actual Service (Google, Yahoo etc.) [7] R. Ukor and A. Carpenter, On modelled flexibility and service
selection optimisation, in 9th Workshop on Business Process
Evaluation Metrics: Modeling, Development and Support, vol. 335.CEUR-WS,
1. No of services Vs the time needed to selecting 2008.
optimal service. [8] D. Ardagna and B. Pernici, Global and local qos guarantee in
web service selection, in Business Process Management Work
2. No of Qos constraints Vs Score value of service. shops, 2005, pp. 3246.
[9] O.Minhyuk, B.Jongmoon,et.al, An efficient approach for QoS-
3. Comparing the meta metrics value Vs no of con- aware service selection based on a Tree-based Algorithm, Seventh
flicts arose of selecting different service for same IEEE international conference on computer and Information
science
activities in different execution paths.
[10] Web Service Business Process Execution Language (WS
BPEL), Version 2.0 - OASIS Committee Draft, 17th May 2006.
[11] Lingshuang S,Lu Z, et.al , Dynamic Availability Estimation For
IJSER 2011
218
ISSN 2229-5518
Service Selection Based On Status Identification, IEEE Interna

tional Conference on web Services, 2008 IEEE
[12] R. Berbner, M. Spahn, N. Repp, O. Heckmann, R.Steinmetz,
Heuristics for QoS-aware web service composition, in: Proc. Int'l
Conf. on Web Services,Sept. 2006.
[13] G. Canfora, M. Di Penta, R. Esposito, M.L. Villani, An ap
proach for QoS-aware service composition based on genetic algo
rithms, in: Proc. Genetic and Computation Conf., June 2005.
[14] Roland Ukor, Andy Carpenter , Flexible Service SelectionOpti
mization Using Meta-metrics Congress on Services-I, 2009 IEEE
[15] D.A. Menasc, V. Dubey, Utility-based QoS brokering in service
oriented architectures, in: IEEE 2007 Intl. Conf. Web Services,
ICWS 2007, Application Services and Industry Track Salt Lake
City, Utah, July 913, 2007, 422430.
[16] D.A. Menasc, H. Ruan, H. Gomma, QoS management in Ser
vice Oriented Architectures, Performance Evaluation Journal 64
(7-8) (2007) 646663.
[17] T. Yu, K.J. Lin, Service selection algorithms for composing complex
services with multiple QoS constraints, in: Proc. of 3rd Int'l Conf.
on Service Oriented Computing, Dec. 2005, pp. 130143.
[18] M. Jaeger, G. Muhl, S. Golze, Qos-aware composition of web se
rvices: A look at selection algorithm, in: Proc. 2005 IEEE Intl.
Conf. Web Services, ICWS05, 2005.
[19] V.Deora,j.Shao,w.A.Gray, supporting qos based selection in ser
vice oriented architecture proceedings of the international confe
rence on next generation web services practices , IEEE 2006
[20] M.A. Serhani, R. Dssouli, A Hafid, H. Sahraoui, A QoS broker
based architecture for efficient web service selection, in: Proc. 2005
IEEE International Conference on Web Services, ICWS05,
2005.
[21] D.Liu, Z.Shao,C.Yu, A heuristic Qos-aware service selection
proach to web service composition, International Conference on
computer and Information science IEEE 2009
IJSER 2011
219

ISSN 2229-5518
Design of Boost Circuit for Wind Generator

N.Prasanna Raj, M.Mohanraj, Rani Thottungal
Abstract This paper presents the design and implementation of a power converter for an autonomous wind induction generator
(IG) feeding an isolated load through the PWM-based novel soft-switching interleaved boost converter. The output voltage and
frequency of the wind IG is inherently variable due to random fluctuation of wind-speed variation. The interleaved boost converter
composed of two shunted elementary boost conversion units and an auxiliary inductor. This converter is able to turn on both the
active power switches at zero voltage to reduce their switching losses and evidently raise the conversion efficiency. Since the two
parallel-operated elem entary boost units are identical, operation analysis and design for the converter module becomes quite
simple. A three-phase induction machine model and a three-phase rectifier-inverter model based on a-b-c reference frame are used
to simulate the performance of the generation system. It can be concluded from the simulated results that the designed power
converters with adequate control scheme can effectively improve the performance of output voltage and frequency of the IG feeding
an isolated load.
Index Terms wind power generator, rectifier-inverter circuit, pulse width modulation (PWM), interleaved boost converter, soft
switching.
method. The third method is the most effective control

1 INTRODUCTION scheme, which can control both voltage and frequency of
the IG within a specified level by using power electronic
T HE characteristics of induction generator (IG) with an
externally connected capacitor bank have been
extensively explored for over 60 years since 1935 [1].
converters such as a rectifier-inverter module (AC-to-DC
and DC-to-AC converters).
Due to the fast development of environmental protection Boost converters are usually applied as preregulators or
concepts, the regenerative or renewable energy sources even integrated with the latter-stage circuits or rectifiers
have been significantly and widely studied and evaluated into single-stage circuits [5 9]. Most renewable power
in the whole world. The primary merits of IG over a sources, such as photovoltaic power systems and fuel cells,
conventional synchronous generator are brushless have quite low-voltage output and require series
construction with squirrel-cage rotor, small size, without connection or a voltage booster to provide enough voltage
DC supply for excitation, less maintenance cost, and better output. Several soft-switching techniques for dc/dc
transient performance. The IGs have been extensively converters have been proposed. The main problem with
utilized as suitable isolated power sources in small these kinds of converters is that the voltage stresses on the
hydroelectric, tidal, and wind energy applications at the power switches are too high in the resonant converters,
remote sites, rural areas, or developing countries [2 4]. especially for the high-input dc-voltage applications.
The performance of IG supplying various static loads using Interleaved boost converters are applied as power-factor-
different control schemes are studied and analysed in correction front ends. An interleaved converter with a
various papers. The control schemes proposed so far coupled winding is proposed to provide a lossless clamp.
divides into three categories. The most economic method is Additional active switches are also appended to provide
by means of switching series and/or parallel capacitors soft-switching characteristics [10 13]. These converters are
connected to the IG stator winding or load side for voltage able to provide higher output power and lower output
regulation. The primary disadvantage of this scheme is that ripple. This paper proposes a soft-switching interleaved
the equivalent capacitance is changed in discrete form and boost converter composed of two shunted elementary
the voltage cannot be effectively and linearly regulated. boost conversion units and an auxiliary inductor. This
Though the voltage magnitude is controlled to a certain converter is able to turn on both the active power switches
constant level, the output frequency of the controlled IG is at zero voltage to reduce their switching losses and
significantly varied with the rotor speed. The second evidently raise the conversion efficiency [14], [15]. Since the
method is to modulate the absorbed reactive power of the two parallel-operated boost units are identical, operation
IG whose stator windings is directly connected to the load analysis and design for the converter module becomes
and the reactive power compensator. Although the quite simple.
terminal voltage of the stator winding can be effectively IG fed to an isolated load through the employment of a
controlled, the frequency variation problem due to random PWM based closed loop boost converter with controlled
fluctuation of rotor speed is similar to the one of the first inverter is presented. The simulated performance of the
IJSER 2011
220

ISSN 2229-5518
proposed control scheme is employed to design a PWM - Pm: Mechanical output power of the turbine (W),
controller for the boost converter and inverter. The - Cp: Performance coefficient of the turbine,
implementation and design of a power converter for an - : Air density (Kg/m3),
autonomous wind induction generator (IG) feeding an - A: Turbine swept area (m2),
isolated load through the PWM-based boost-inverter circuit - V: Wind speed (m/s).
is simulated here. Pulse-width modulation (PWM) is a very efficient way
of providing intermediate amounts of electrical power
between fully on and fully off. A simple power switch with
2 METHODOLOGY a typical power source provides full power only when
The generation system is designed with IG. The stator switched on. The term duty cycle describes the proportion
winding terminals of the IG are connected to the load of on time to the regular interval or period of time; a low
through the rectifier, DC link, Boost circuit and inverter. duty cycle corresponds to low power, because the power is
The closed loop PWM signal generates proper PWM off for most of the time. Pulse-width modulation uses
signals to switch the 2 power electronics devices of the a rectangular pulse wave whose pulse width is modulated
Interleaved Boost Converter (IBC). The wind turbine resulting in the variation of the average value of the
rotates the IG. The IG generates power when the speed of waveform. If we consider a pulse waveform f(t) with a low
the turbine is above the rated speed. The power generated value ymin, a high value ymaxand a duty cycle D, the
from the IG is converted to DC with a diode bridge average value of the waveform is given by
rectifier. The obtained DC voltage will not be in a pure DC
signal. A filter circuit is used to filter out the ripple current
and a pure DC voltage is obtained. This DC voltage is then As f(t) is a pulse wave, its value is ymax for 0<t<D.T
boosted to the required DC level and then converted to and ymin for D.T<t<T The above expression then becomes
three phase AC signal with IGBT which is driven by PWM
signal. To regulate the AC output voltage the IBC is
controlled by close loop PWM signals. A load is connected
at the output of the inverter. The voltage-current equations
of the studied IG in matrix form are listed as below.
The simplest way to generate a PWM signal is the

Where Rs(abc) and R'r(abc) are respectively the interceptive method, which requires only a sawtooth or a
resistance matrices of the stator and rotor windings and triangle waveform and a comparator. When the value of
s(abc) and 'r(abc) are respectively the flux-linkage the reference signal is more than the modulation
matrices of the stator and rotor windings [16]. waveform, the PWM signal is in the high state, otherwise it
The expression of the output power is is in the low state.
Pm=0.5AV3Cp
Convert L Inverter
IG
IBC Load
Cdc
Wr
Wind Turbine
Ce PWM
PWM Gating Signal
Driver
DSPIC
Fig1 Block Diagram
IJSER 2011
221

ISSN 2229-5518
Fig 2 Interleaved Boost Converter
Fig.2 shows the proposed soft-switching converter

module. Inductor L1, MOSFET active switch S1 and diode
D1 comprise one step-up conversion unit, while the
components with subscript 2 form the other. Dsx and Csx
are the intrinsic antiparallel diode and output capacitance
of MOSFET Sx, respectively. The voltage source Vin, via
the two paralleled converters, replenishes output capacitor
Fig 3 IBC Study
C0 and the load. Inductor Ls is shunted with the two active

MOSFET switches to release the electric charge stored
within the output capacitor Csx prior to the turn-ON of Sx
to fulfill zero-voltage turn- ON (ZVS), and therefore, raises
the converter efficiency. To simplify the analysis, L1, L2
and C0 are replaced by current and voltage sources,
respectively, as shown in Fig. 3.
The proposed circuit is focused on higher power
demand applications. The inductors L1 and L2 are likely to
operate under continuous conduction mode (CCM);
therefore, the peak inductor current can be alleviated along
with less conduction losses on active switches. Under CCM
operation, the inductances of L1 and L2 are related only to
the current ripple specification. What dominates the output
power range and ZVS operation is the inductance of Ls.
3 CIRCUIT OPERATION ANALYSIS

Before analysis on the circuit, the following assumptions Fig 4. Modes of Operation of IBC
are presumed.
neglect the output voltage ripple.
1) The output capacitor C0 is large enough to reasonably
IJSER 2011
222

ISSN 2229-5518
2) The forward voltage drops on MOSFET S1, S2 and three-phase rectifier-inverter model based on a-b-c
diodes D1 and D2 are neglected. reference frame to simulate the performance of the
3) Inductors L1 and L2 have large inductance, and their generation system. The IG design is made with the
currents are identical constants, i.e., IL1=IL2=IL. calculated value of resistance, flux linkage of the stator and
4) Output capacitances of switches Cs1 and Cs2 have the rotor windings and with the torque equation and the
same values, i.e., Cs1=Cs2=Cs. number of poles. The generated power is rectified. The
The two active switches S1 and S2 are operated with generated power varies with the wind speed. The power
pulsewidth-modulation (PWM) control signals. They are converter converts the three phase AC to DC and then the
gated with identical frequencies and duty ratios. The rising filter circuit is used for obtaining smooth DC voltage across
edges of the two gating signals are separated apart for half it. This DC voltage is then boosted with the interleaved
a switching cycle. The operation of the converter can be boost converter (IBC). Theoretically IBC can boost up to
divided into eight modes, and the equivalent circuits and 200%. The DC voltage is fed to the inverter before which it
theoretical waveforms are illustrated in Fig. 4. is regulated to desired voltage in a closed loop with PWM
The project module is simulated using MATLAB technique. Thus the output at the IBC is always constant
7.7.0(R2008b). The simulation is executed under ode23tb rated voltage. This DC voltage is then regulated AC three
(stiff/TR-BDF2) state which is used to fasten the execution phase with IGBTs. An IGBT inverter is designed with a
speed and the Zero-crossing control is disabled. The solver open loop PWM technique and feed the AC load. PWD
method is set to fast. The voltage is measured at different signal is designed in close loop and open loop system with
points in the simulation circuit. The simulated output is PID and PI controller respectively. The PWM signal is
shown below. The system is tested with different load and generated to regulate the DC voltage that in turn gives a
wind speed. The designed system generates AC power fixed AC supply from the inverter. Thus a regulated Three
with asynchronous generator (215HP; 400V; 50Hz). The Phase AC supply is transmitted to the Load.
utilizes a three-phase asynchronous machine model and a
4 RESULTS AND DISCUSSIONS
Fig 6 Input current 10A

Fig 5 Input voltage 160V AC
Fig 8 IBC controlled Output 220V DC

Fig 7 Rectified DC 160V
IJSER 2011
223

ISSN 2229-5518
Fig 10 Output current 3A (IBC)

Fig 9 Output Voltage with IBC 220V AC
Fig 12 THD for input voltage 160V AC

Fig 11 Output Voltage without IBC 120V
shown in fig 8. The boosted voltage is converted to Three phase

AC with the inverter and the voltage obtained is the required
output voltage level of 220V and the current is 3A, shown in fig
9 & 10. The output without IBC is shown in fig 11, the obtained
voltage will be same the voltage generated at input. The THD
value is calculated for the input voltage and the output voltage.
The graph shows that the THD has decreased compared to the
input voltage, shows in fig 12 & 13.
5 CONCLUSION
Thus from the above simulated results we can conclude
at that even at low wind speed and low power generation
the required voltage can be obtained at the output. The
main advantage of the system is that the minimum
Fig 13 THD for output Voltage 220V AC required wind speed for the power generation can be
reduced and generator power ratings can be reduced. The
The input voltage and current generated at the wind induction gear mechanism between the turbine and the shaft can be
generator is shown in fig 5 & 6. The generated voltage will vary reduced. The displayed out is taken when the wind speed
for different wind speed. The generated voltage is noted for is 7m/s. At this speed the generator generates 160V AC
different wind velocity and at each voltage level the output and this voltage is boosted in the IBC to 220V DC which is
voltage is noted. The examined result shows that the output the required voltage level. Steady-state results under
voltage is maintained at 220V even when the generated voltage various loads show that the designed control system for IG
is below the desired voltage level. The simulated result shown can maintain at desired levels. The simulated results of
here is for generated voltage of 160V and current 9A. The output voltage validate the required performance of the
generated voltage is converted to DC 160V by the rectifier, proposed control scheme. Thus the voltage obtained at the
shown in fig 7. This DC voltage is boosted in the closed loop output is regulated with the IBC and inverter block.
circuit with the IBC (Interleaved Boost Converted) to 220V DC,
IJSER 2011
224

ISSN 2229-5518
REFERENCES Jan. 2008.

[17] W. Li and X. He, ZVT interleaved boost converters for high-ef-
[1] E.D. Basset and F.M. Potter, Capacitive excitation of induction
ciency, high step-up DCDC conversion, IET Electron. Power
generators, Trans. American Institute Electrical Engineering, vol.
Appl.,vol. 1, no. 2, pp. 284290, Mar. 2007.
54, pp. 540-545, 1935.
[18] W. Li and X. He, An interleaved winding-coupled boost
[2] J. Arrilaga and D.B. Watson, Static power conversion from self-
converter with passive lossless clamp circuits, IEEE Trans. Power
excited induction generators, IEE Proceedings, vol. 125, no. 8, pp.
Electron., vol. 22, no. 4, pp. 14991507, Jul. 2007.
743-746, 1978.
[3] R. Ramakumar, Renewable energy sources and developing
countries, IEEE Trans. Power Apparatus and Systems, vol. 102, no.
2, pp. 502-510, 1983
[4] S.S. Murthy, O.P. Malik, and A.K. Tandon, Analysis of self
excited induction generators, IEE Proceedings, Part C BIOGRAPHIES
Generation, Transmission, and Distribution, vol. 129, no. 6, pp. 260-
265, 1982. N.Prasanna Raj, from Chennai, finished his UG in Coimbatore Institute of
Engineering and Information Technology and pursuing his final year M.E. in
[5] B. Singh and L.B. Shilpakar, Analysis of a novel solid state
Power Electronics and Drives in Kumaraguru College of Technology,
voltage regulator for a self-excited induction generators, IEE Coimbatore. ([email protected])
Proceedings, Generation Transmission and Distribution, vol. 145, no.
6, pp. 647-655, 1998.
M.Mohanraj, from Erode, finished his UG in Bharathiar University and he
[6] E.G. Marra and J.A. Pomilio, Induction-generator-based system
obtained PG in Power System Engineering in Annamalai University and he is
providing regulated voltage with constant frequency, IEEE currently working as Assistant Professor in EEE department, Kumaraguru
Trans. Industrial Electronics, vol. 47, no. 4, pp. 908-914, 2000. College of Technology, Coimbatore and a Life member of ISTE. His research
[7] E.G. Marra and J.A. Pomilio, Self-excited induction generator area includes Wind Energy Conversion, Solar Energy, Machines and Power
Quality Issues.([email protected])
controlled by a VS-PWM bidirectional converter for rural
applications, IEEE Trans. Industry Applications, vol. 35, no. 4, pp.
877-883, 1999. Rani Thottungal, obtained her B.E and M.E degrees from Andhra University
[8] S. Wekhande and V. Agarwal, Simple control for a wind-driven and Doctorate from Bharathiar University. She is currently working as
Professor and Head in EEE department, Kumaraguru College of Technology,
induction generator, IEEE Industry Applications Magazine, vol. 7,
Coimbatore. Her research interest includes Power System, Power Inverter and
no. 2, pp. 45-53, 1999. Power Quality Issues.
[9] C. M.Wang, A new single-phase ZCS-PWM boost rectifier with
high power factor and low conduction losses, IEEE Trans. Ind.
Electron., vol. 53, no. 2, pp. 500510, Apr. 2006.
[10] Y. Jang, D. L. Dillman, and M. M. Jovanovic, A new soft-
switched PFC boost rectifier with integrated flyback converter for
stand-by power, IEEE Trans. Power Electron., vol. 21, no. 1, pp.
6672 , Jan. 2006.
[11] Y. Jang, M. M. Jovanovic, K. H. Fang, and Y. M. Chang, High-
powerfactor soft-switched boost converter, IEEE Trans. Power
Electron., vol. 21, no. 1, pp. 98104, Jan. 2006.
[12] K. P. Louganski and J. S. Lai, Current phase lead compensation
in single-phase PFC boost converters with a reduced switching
frequency to line frequency ratio, IEEE Trans. Power Electron.,
vol. 22 , no. 1, pp. 113119, Jan. 2007.
[13] K. Kobayashi, H. Matsuo, and Y. Sekine, Novel solar-cell power
supply system using a multiple-input DCDC converter, IEEE
Trans. Ind. Electron., vol. 53, no. 1, pp. 281286, Feb. 2006.
[14] S. K. Mazumder , R. K. Burra, and K. Acharya, A ripple-
mitigating and energy-efficient fuel cell power-conditioning
system, IEEE Trans. Power Electron. , vol. 22, no. 4, pp. 14371452,
Jul. 2007.
[15] G. Yao, A. Chen, and X. He, Soft switching circuit for
interleaved boost converters, IEEE Trans. Power Electron., vol. 22,
no. 1, pp. 8086, Jan. 2007.
[16] Q. Ting and B. Lehman, Dual interleaved active-clamp forward
with automatic charge balance regulation for high input voltage
application , IEEE Trans. Power Electron., vol. 23, no. 1, pp. 3844,
IJSER 2011
225

ISSN 2229-5518
A Novel Method for Fingerprint Core Point

Detection
Navrit Kaur Johal, Prof. Amit Kamra
Abstract- Fingerprint recognition is a method of biometric authentication that uses pattern recognition techniques based on high-
resolution fingerprints images of the individual. Fingerprints have been used in forensic as well as commercial applications for
identification as well as verification. Singular point detection is the most important task of fingerprint image classification operation.
Two types of singular points called core and delta points are claimed to be enough to classify the fingerprints. The classification can
act as an important indexing mechanism for large fingerprint databases which can reduce the query time and the computational
complexity. Usually fingerprint images have noisy background and the local orientation field also changes very rapidly in the singular
point area. It is difficult to locate the singular point precisely. There already exists many singular point detection algorithms, Most of
them can efficiently detect the core point when the image quality is fine, but when the image quality is poor, the efficiency of the
algorithm degrades rapidly. In the present work, a new method of detection and localization of core points in a fingerprint image is
proposed.
Index TermsCore Point, Delta Point, Smoothening, Orientation Field, Fingerprint Classes

Core points are the points where the innermost ridge loops
1. INTRODUCTION are at their steepest. Delta points are the points from which
ingerprints have been used as a method of identifying three patterns i.e. loop, delta and whorl deviate. Definitions
F individuals due to the favorable characteristics such as may vary in deferent literatures, but this definition of
unchangeability and uniqueness in an individuals singular point is the most popular one. Figure 1 below
lifetime. In recent years, as the importance of information represents the core and delta points.
security is highly demanded, fingerprints are utilized for
the applications related to user identification and
authentication. Most Automatic Fingerprint Identification Core Point
systems are based on local ridge features; ridge ending and
ridge bifurcation, known as minutiae The first scientific
study of the fingerprint was made by Galton who divided
fingerprint into three major classes: arches, loops, and
whorls. Henry, later refined Galtons classification by
increasing the number of classification. Henrys
classification is well-known and widely accepted. Henrys
classes consist of: arch, tent arch, left loop, right loop and Delta Points
whorl .
At a global level the fingerprint pattern exhibits the area Fig 1. The Core and Delta Points on a fingerprint image
that ridge lines assume distinctive shapes. Such an area or
region with unique pattern of curvature, bifurcation, This paper is organized as follows. In section 2, are
termination is known as a singular region and is classified discussed the different types of fingerprints. In Section 3 is
into core point and delta point. The singular points can be explained the drawbacks with the existing techniques of
viewed as the points where the orientation field is core point detection. Section 4 focuses on the problem
solution. In section 5, the core point is extracted using the
discontinuous. proposed algorithm. The experimental results performed on
a variety of fingerprint images are discussed in section 6
and the conclusion and future scope is discussed in Section

Navrit Kaur is pursuing M.Tech from Guru Nanak Dev Engineering
7.
College Ludhiana E-mail: [email protected]
Amit Kamra is with Guru Nanak Dev Engineering College, Ludhiana. E-
mail: [email protected].
IJSER 2011
226

ISSN 2229-5518
2 FINGERPRINT CLASSES
The positions of cores and deltas are claimed to be enough
to classify the fingerprints into six categories, which include
arch, tented arch, left-loop, right-loop, whorl, and twin-
loop.
Loops constitute between 60 and 70 per cent of the
patterns encountered. In a loop pattern, one or
more of the ridges enters on either side of the
impression, recurves, touches or crosses the line of
the glass running from the delta to the core, and
terminates or tends to terminate on or in the
direction of the side where the ridge or ridges
entered. There is exactly one delta in a loop. Loops Fig 2. Classes of fingerprint (a)Arch, (b)Tented Arch, (c) Right Loop (d)
Left Loop, (e) Whorl and (f) Double Loop (The double loop type is
that have ridges that enter and leave from left side sometimes counted as whorl)
are called the Left Loops and loops that have ridges
that enter and leave from right side are called the
Fingerprint friction ridge details are generally described in
Right Loops. In twin loops the ridges containing
a hierarchical order at three levels, namely, Level 1
the core points have their exits on different sides.
(pattern), Level 2 (minutiae points) and Level 3 (pores and
ridge shape).Automated fingerprint identification systems
In a whorl, some of the ridges make a turn through
(AFISs) employ only Level 1 and Level 2 features. No two
at least one circuit. Any fingerprint pattern which
fingerprints are alike, but the pattern of our fingerprint is
contains 2 or more deltas will be a whorl pattern
inherited from close relatives and people in our immediate
family. This is considered "level 1 detail." The detail of our
In arch patterns, the ridges run from one side to the
actual finger and palm print is not inherited. This is
other of the pattern, making no backward turn.
considered "level 2 and 3 level detail" and is used to identify
Arches come in two types, plain or tented. While
fingerprints from person to person.
the plain arch tends to flow rather easily through
The following figure briefly explains the three types of
the pattern with no significant changes, the tented
levels of details in our fingerprint:
arch does make a significant change and does not
have the same easy flow that the plain arch does.
In short, while classifying the fingerprints, we can make the

assumption that if a pattern contains no delta then it is an
arch, if it contains one (and only one) delta it will be a loop
and if it contains 2 or more it will always be a whorl. If a
pattern does contain more than 2 deltas it will always be an
accidental whorl.
Fig 3. Level 1 , Level 2 and Level 3 Details
In this paper, we propose a new core point detection

method which can precisely localize the core point and does
not require further post processing as well. The proposed
method only concentrates on the core point detection as
most of the ridge characteristics e.g ridge endings and ridge
bifurcations are present in the core block (centre).
3 PROBLEM FORMULATION
The existing techniques used for detection of core point do
not produce good results for noisy images. Moreover, they
IJSER 2011
227

ISSN 2229-5518
may sometimes detect spurious core point due to the a) Firstly, the image I(i,j) is divided into non
inability to work efficiently for noisy images. Also overlapping blocks of size wXw.
techniques like Poincare Index fail for Arch type of Image. b) The mean value M(I) is then calculated for each
The aim of proposed algorithm is to formulate a more block using the following equation:
accurate core point determination algorithm which can w w
2 2
produce better localization of core points avoiding any M ( I ) 12 I (i, j ) (1)
spurious points detected producing robust results for all w iw j w
2 2
types of fingerprints that have been discussed in this paper.
c) The mean value calculated above is then used to
find the variance using the following equation:
4 PROPOSED SOLUTION w w
2 2
V ( I ) 12 (I (i, j) M (I )) 2
(2)
w iw j w
Original Image 2 2
d) If the variance is less than the global threshold

value selected empirically, then the block is
Normalized Image
assigned to be a background region; otherwise, it is
assigned to be part of the foreground.
Smoothed Image
Fine Tuning of
Orientation Field
Fig 5. The result of segmentation using a variance threshold of 100 and

Core Point on the a block size of 16 16.
Smoothed Image
4.2 NORMALIZATION
In image processing, normalization is a process that
Fig 4. The proposed methodology for Core Point Detection
changes the range of pixel intensity values. Normalization
4.1 SEGMENTATION is sometimes called contrast stretching. In more general
The first step of the fingerprint enhancement algorithm is fields of data processing, such as digital signal processing, it
image segmentation. Segmentation is the process of is referred to as dynamic range expansion. The purpose of
separating the foreground regions in the image from the dynamic range expansion in the various applications is
background regions. The foreground regions correspond to usually to bring the image, or other type of signal, into a
the clear fingerprint area containing the ridges and valleys, range that is more familiar or normal to the senses, hence
which is the area of interest. The background corresponds the term normalization. Normalization is a linear process. If
to the regions outside the borders of the fingerprint area, the intensity range of the image is 50 to 180 and the desired
which do not contain any valid fingerprint information. range is 0 to 255 the process entails subtracting 50 from each
Cutting or cropping out the region that does not contain of pixel intensity, making the range 0 to 130. Then each
valid information minimizes the number of operations on pixel intensity is multiplied by 255/130, making the range 0
fingerprint image. The background regions of a fingerprint to 255.
image generally exhibit a very low grey-scale variance Let I(i,j) denote the gray-level value at pixel (i,j), M0 and V0
value, whereas the foreground regions have a very high denote the estimated mean and variance of I, respectively,
variance. Hence, a method based on variance threshold can and N(i,j) denote the normalized gray-level value at pixel (i,
be used to perform the segmentation. The steps for mean j). The normalized image is defined as follows:
and variance based segmentation are as follows:
IJSER 2011
228

ISSN 2229-5518
M V0 ( I (i , j ) M i ) 2 if I (i, j ) M Ridge smoothing is then performed which is a process of

N (i , j ) 0 (3) finding out the valid frequency of the binary image of
2 otherwise
M 0 V0 ( I (i, j ) M i ) ridges. Filters corresponding to these distinct frequencies
and orientations are then generated. Fig 7 indicates the
Normalization is a pixel-wise operation. It does not change results obtained after smoothing the image.
the clarity of the ridge and valley structures. The main
purpose of normalization is to reduce the variations in gray-
level values along ridges and valleys, which facilitates the
subsequent processing steps.
Fig 7. Original and Smoothed Fingerprint images respectively.
The direction of gravity of progressive blocks is then

determined, using the following equations (P=3):
Fig 6. Result of normalization.(a) Input image(b) Normalized (M0 = 100,
VAR0 = 100).
P 1 P 1 P 1 P 1
A V x and B V y (7)
4.3 ORIENTATION ESTIMATION k 0 l 0 k 0 l 0
The orientation flow is then estimated using the least square As we know that singular points are the points where the
method using the following equations after dividing the orientation field is discontinuous, hence orientation plays a
input image I into non overlapping blocks of size wXw and crucial role in estimating the core point on a fingerprint
then computing the gradients x and y at each pixel. image. Hence, we need another mechanism to fine tune the
i w/ 2 j w/ 2 orientation field so as to avoid any spurious core points and
Vx(i, j) 2x(u, v) y(u, v) (4) the irregularities that has occurred because of noise. The
u i w / 2 v j w / 2 orientation field for coarse core point is then fine tuned by
adjusting orientation using the following :
i w/ 2 j w/ 2
Vy(i, j) 2 x(u, v)2 y(u, v) (5) If : B( i , j ) 0 then: 0.5 tan 1 ( B / A) (8)
u i w / 2v j w / 2
else: / 2
Where x (u,v) and y (u,v) represents gradient if 0 then
magnitudes at each pixel in x and y directions respectively. if : A 0 then: / 2
else:
The direction of block centered at pixel (i,j) is then else if A 0 then: / 2
computed using the following equation:
V y (i, j ) Hence we calculate the value , which is the orientation
( i , j ) 1 tan 1
(6) value of the image.
2 V x ( i. j )
Due to the presence of noise, corrupted ridge and valley 4 PROPOSED ALGORITHM
structures, minutiae etc. in the input image, the estimated
1. The original fingerprint image is first segmented
local ridge orientation, (i,j), may not always be correct. A
and normalized using equations (1), (2) and (3).
low-pass filter is hence used to modify the incorrect local
ridge orientation. 2. Determine the x and y magnitudes of the gradients
Gx and Gy at each pixel.
4.4 SMOOTHING AND FINETUNING
IJSER 2011
229

ISSN 2229-5518
3. Divide the input image into blocks of size wXw.
4. Estimate the local orientation using equations

(4),(5),(6) and perform ridge smoothening.
5. Of the wXw image size, determine the direction of

gravity of progressive blocks (non-overlapping sub
block ) using equation (7).
6. Fine tune the orientation field.
7. The blocks with slope values ranging to 0 to pi/2

are located. Then a path is traced down until we
encounter a slope that is not ranging from 0 to pi/2
and that block is marked.
8. The block that has the highest number of marks

will compute the slope in negative y direction and
output on x and y position which will be the core
point. The proposed algorithm gives us near accurate results for
noisy images as shown below:
5 RESULTS AND DISCUSSION

The proposed technique can localize the fingerprints at a
good success. Experiments have been performed on nearly
180 fingerprints from different fingers and a variation in
noise has been taken into consideration.
The results have been obtained in MATLAB, the core point
positions hence obtained are used to narrow down the
search when using a huge database of fingerprint images in Fig 10. Accuracy of the proposed technique for a poor input
applications like biometrics security, forensics etc. The image.
following table clearly compares the accuracy of results of
the proposed technique with the existing state of the art But the algorithm failed its core point detection for images
technique. having a very poor quality as shown in figure 11. Those
fingerprints are too oily or wrinkled.
TABLE 1: Comparison of results of the proposed technique
with the existing technique.
Original Image Proposed Existing

Technique Technique
Fig 11. Poor quality fingerprint images
6 CONCLUSION AND FUTURE SCOPE

This paper proposes a novel method to consistently and
precisely locate the singular points in fingerprint images.
The method applied is based on the enhanced fingerprint
IJSER 2011
230

ISSN 2229-5518
image orientation reliability. Our future work will focus on

improvements in locating the secondary singular points of
fingerprint images and classifying the fingerprints based
upon the location of the singularities so that the
computational time for matching fingerprints in a huge
database may reduce.
7 REFERENCES
[1] Keokanlaya Sihalath, Somsak Choomchuay, and Kazuhiko Hamamoto,

Core Point Identification with Local Enhancement, JCSSE, May 13-15, 2009
(Vol. 1)
[2] Mohammed S. Khalil, Dzulkifli Muhammad, Muhammad K. Khan and
Khaled Alghathbar, Singular Point Detection using Fingerprint Orientation
Field Reliability , International Journal of Physical Sciences Vol. 5(4), pp. 352-
357, April 2010
[3] Wang Feng, Chen Yun , Wang Hao, Wang Xiu-you Fingerprint
Classification Based on Improved Singular Points Detection and Central
Symmetrical Axis , 2009 International Conference on Artificial Intelligence
and Computational Intelligence
[4] Filipe Magalhes, Hlder P. Oliveira, Aurlio C. Campilho, A New
Method for the Detection of Singular Points in Fingerprint Images , 978-1-
4244-5498-3/092009 IEEE
[5] Galton F., Fingerprint, McMillan, London, 1892.
[6] Henry E., Classification and uses of fingerprints , Rout ledge, London,
1900.
[7] Weiwei Zhang, Yangsheng Wang , Core-Based Structure Matching
Algorithm of Fingerprint Verification , National Laboratory of Pattern
Recognition, Institute of Automation,Chinese Academy of Sciences
[8] Tomohiko Ohtsuka, Daisuke Watanabe, Hiroyuki Aoki1, Fingerprint
Core And Delta Detection By Candidate Analysis , MVA2007 IAPR
Conference on Machine Vision Applications, May 16-18, 2007, Tokyo, Japan
[9] S.Chikkerur, C. Wu and Govindaraju, A Systematic Approach for
Feature Extraction in Fingerprint Images, ICBA 2004.
[10] L. Hong, Y. Wan, and A. Jain, Fingerprint Enhancements: Algorithm
and Evaluation , Proc. IEEE Trans. on Pattern Analysis and Machine
Intelligence, vol.20, no. 8, pp. 777-789, 1998.
[11] M. Liu, X. Jiang, and A. C. Kot, Fingerprint Reference Point Detection ,
Journal of Applied Signal Processing, vol. 4, pp. 498509, 2005.
[12] C.-Y. Huang, L. Liu, and D. D. Hung, Fingerprint Analysis and Singular
Point Detection.
[13]Ramandeep Kaur, Parvinder S. Sandhu, Amit Kamra,A Novel Method
for Fingerprint Feature Extraction 2010 International Conference on
Networking and Information Technology 978-1-4244-7578-0 2010 IEEE .
[14] Bob Hastings, A Statistical Investigation of Fingerprint Patterns
Department of Computer Science & Software Engineering The University of
Western Australia, Australia.
Navrit Kaur Johal received her B.Tech degree in Computer Science

and Engineering from AIET, Faridkot and is pursuing her M.Tech in
Computer Sc. and Engg from Guru Nanak Dev Engg. College,
Ludhiana.
Amit Kamra received his B.Tech degree in Electronics Engineering in

2001. and completed his M.Tech in Information Technology in 2004.
He is working as Associate Professor in Guru Nanak Dev Engg.
College, Ludhiana and is currently pursuing his Ph.D in image
processing.
IJSER 2011
231

ISSN 2229-5518
Robust Impulse Eliminating Internal Model

Control of Singular Systems: A Robust Control
Approach
M.M. Share Pasand, H.D. Taghirad
Abstract The problem of model based internal control of singular systems is investigated and the limitations of directly extending
the control schem es for normal systems to singular ones were analyzed in this paper. A robust approach is proposed in order to
establish the control scheme for singular systems, and moreover, to present a fram ework for robust control of singular systems in
presence of modeling uncertainties. The theory is developed through a number of theorems, and several simulation examples are
included and their physical inter-pretations are given to verify the proof of concept.
Index Terms singular systems, Impulsive behavior, internal model control, Model based control, robust control, tracking problem,
impulse elimination.
ing framework for analyzing the algebraic loop, regularity of

1. INTRODUCTION the closed loop and impulse elimination problems of singular
S ingular systems represent a general framework for linear

systems [1]. A singular model is an appropriate model for
describing large scale interconnected systems, constrained ro-
systems. Furthermore, most of existing methods in robust con-
trol of singular systems are limited to study a special form of
uncertainty. They assumed matrix E to be exactly known [6, 7
bots and other differential algebraic systems with linear alge- and 10]. This assumption is more restrictive than it appears,
braic constraints [2]. Also singular models can be utilized to because it limits the system to be impulsive while some uncer-
model a system when the dependent variable is displacement tainties may exist which lead to a strictly proper system for a
rather than time [3]. Since the first time they were introduced singular model. Therefore this paper suggests a new concept
[4], several efforts have been made to control singular systems for robust control. While previous works on robust control fo-
[5-9]. As the singular systems were firstly introduced in the cuses on robust stability and robust performance, as it comes to
state space form representation [4], they were usually studied in descriptor systems, robust properness of closed loop should be
time domain. In [5] the problem of finite mode pole placement studied. The internal model framework for controlling singular
is studied, while simultaneous impulse elimination and robust systems provides a more logical uncertainty model and release
stabilization problem is considered in [6], robust Eigen- the restrictive assumptions made in the existing state space
structure assignment of finite modes is studied in [7]. In [8] methods for robust control of singular systems. Also it provides
strict impulse elimination is studied using time derivative feed- offset free tracking capability of the closed loop as well as be-
back of the states and [9] investigated the output feedback con- ing able to well treat delayed systems. The main obstacle which
trol using a compensator. In fact most of the existing methods arises in the internal model control of singular systems is that
are extensions of the control schemes for standard systems the internal model cannot be implemeted easily, because it is
[5],[6],[10]. In the singular system control context the control generally improper. Even in computer aided control systems it
objectives are more complicated due to the existing obstacles is not easy to simulate a singular system, since the discrete
such as algebraic loop phenomenon, impulsive behavior [11] , model needs future input data to determine the system state
and regularity of the closed loop [8,9] . Unlike the time domain vector at the present time [1]. This problem results in an inevit-
methods, there are very few works on the frequency domain able mismatch between the plant and the parallel model used in
control of singular systems. In the frequency domain, the track- IMC.
ing problem, robust control problem and impulse elimination
Notice that, defining the disk shaped multiplicative uncer-
can be treated more conveniently. Specifically the so called
tainty leads to an unbounded uncertainty profile which is not
Internal Model Control (IMC) method provides a very interest-
suitable in robust design of singular systems. This paper pro-
vides solution to the latter problem by introducing the singular

internal model filter in series with the conventional internal
M.M. Share Pasand was graduated from KNT University of technology.
He is currently within IAU Malard branch.
model filter. The aforementioned filter eases the design proce-
E-mail: [email protected]. dure, bounds the uncertainty profile. Also it makes the closed
Dr. H.D. Taghirad is a professor within the control systems group, KNT loop strictly proper and eliminates impulsive modes by smooth-
University of technology. ing the control action as much as needed. Another role of the
introduced filter is to make it possible to design robust control-
IJSER 2011
232

ISSN 2229-5518
ler in the conventional context. The paper is organized as fol- Corollary1: A singular system is called impulse free, if and
lows. In the next section backgrounds are discussed and the only if, it doesnt exhibit impulses in its impulse response.
obstacles in control of singular systems are presented, and
Definition2: A singular system is called minimal if it is observ-
some major limitations of the direct extension of IMC are ex-
able and controllable. The minimality of the plant is presumed
plained. In the third section the proposed method is studied and
throughout this paper.
the filter design procedure is illustrated. In the fourth section
several examples and simulations are given to examine the al- Definition3: A transfer function is strictly proper, bi-proper and
gorithm both in terms of robustness properties and closed loop improper if the following limit is zero, a finite nonzero value
performance. Finally, the concluding remarks are given in last and infinite respectively. Strictly proper and bi-proper systems
section. may be generally named as proper.
lim ( s)
CONTROL OBJECTIVES IN SINGUALR CONTROL s
SYSTEMS Lemma1: A state space realization of a singular system is im-

pulse free if and only if its transfer function has a nonnegative
Definitions and Singular Systems Characteristics relative degree i.e. it is proper.
As Descriptor models are a straight extension of standard
Proof: The transfer function matrix from input to state for sys-
state space models [1], control problem for these systems has a
tem (1) can be computed as:
wider range of objectives. A control system for a standard plant
is designed such that the closed loop is stable and has a prede- C adj ( sE A) B
(s ) C (sE A) 1 B
fined performance and acceptable robustness properties. A sin- sE A
gular control system, on the other hand, should be designed
such that it is impulse free, regular and doesnt include any It is known that degree of the nominator is equal to rank of E at
algebraic loops in addition to the aforementioned properties. most, therefore if condition (2) is satisfied then the system
These control objectives combined with the standard objectives transfer matrix will be proper and if not, the transfer matrix
make the control of singular systems more challenging. Robust may be improper. On the other hand if transfer matrix is not
control of singular systems is the most challenging issue be- proper condition (2) is not satisfied.
cause it requires robustness not only in the stability and per- Remark2: The observability assumption is essential for the
formance but also in regularity and properness. State space above lemma .It can be shown that it may be a number of on-
robust control schemes require robust observers in order to observable impulsive modes which do not appear in the output.
work properly and do not guarantee strict properness of the Also note that condition (2) is a general condition for impulse
closed loop also they usually result in more complicated deriva- free systems but in order to equate it with corollary 1, the ob-
tion algorithms. The main advantage to use internal model con- servability assumption is needed.
trol scheme in here, is that the IMC provides an effective tool
in frequency domain without introducing complicated methods Lemma2: In the unity output feedback structure the closed loop
in evaluation of closed loop performance and stability. There- system is strictly proper if the compensator/plant combination
fore, IMC can be regarded as a proper alternative for existing is strictly proper.
state space methods. Also IMC provides a simple framework Proof: Expand the nominator and denominator by their respec-
for algebraic loop and properness analysis of singular control tive Taylor series.
systems which is much simpler than that in state space methods
or other frequency domain schemes. Consider the following a k s k
state space description: (s) k 1

Ex Ax Bu 1 b j sj
(1) j 1
y Cx
Because CP is supposed to be strictly proper, the largest term in
Definition1: System (1) is impulse free if and only if: its expansion has a negative power therefore the denominator
has a greater degree than the nominator and thus the closed
deg sE A rank E (2)
loop system is strictly proper.
The nullity index of E is called singularity index of a singular
system (1) in this paper.
Remark1: Note that the following general inequality always
holds:
Figure1: Feedback structure
deg sE A rank( E ) (3)
IJSER 2011
233

ISSN 2229-5518
Remark3: Note that Lemma2 provides a sufficient condition. certainty descriptions are limited to represent only singular
The necessary and sufficient condition is derived later. Lem- systems with a pre-specified singularity index. If one aug-
ma2 shows that why the objective of properness has not been ments the improper plant by high frequency stable poles a
considered before the introduction of descriptor systems. As- strictly proper model can be obtained, which has a very
suming strictly proper functions for plant and compensator, it is close behavior to plant at least at low enough frequency
trivial that the closed loop system is strictly proper. Also for a range. Larger poles results in closer response to that of the
strictly proper plant and a bi-proper compensator the closed plant in wider bandwidths. However, in this way the uncer-
loop will be bi-proper. tainty becomes unbounded. In particular assume a poly-
Lemma3: For a bi-proper plant and a bi-proper compensator, nomial of stable real poles with a unit steady state gain
the closed loop will be improper if and only if: namely D, then one can write:
lim CP 1 ~ p( s)
s p ( s) (5)
D( s )
Lemma4: In order to have a strictly proper closed loop system
The above description for model is the most natural selec-
with a unit feedback, if the plant is improper the compensator
should be strictly proper with a sufficiently large relative de- tion for a strictly proper model, whose behavior is as close
gree. as possible to that of the plant. However, in this situation
the mismatch between plant and model is not included in a
Proof: For the closed loop to be strictly proper, the compensa- disk shaped region. In other words the uncertainty bound
tor/plant should be strictly proper according to lemma2, there- will be infinity. Now we can take different approaches:
fore the compensator should be strictly proper. Choose another internal model which yields bounded un-
certainty; developing new theory for this kind of uncertain-
Robust Internal Model Control of Singular Systems
ty; or modify the plant input in order to bind the uncertain-
In IMC structure the parallel model or internal model is ty as well as removing impulses from the response. The
inevitably proper or strictly proper. Therefore there always following lemmas are introductory materials for the theo-
exists a mismatch between the plant and model. For a con- rems developed later in this paper.
tinuous output especially in case of initial jumps of the in-
put, it is required that the plant and the internal model have Lemma5: A control system is robustly stale, if and only if,
the same infinite gain and the compensator is strictly prop- the complementary sensitivity function fulfills the follow-
er. This issue can be treated by a smoothing pre-filter for ing inequality: [12]
reference signal but the method is not robust against model
sup ( s)lm 1 (6)
uncertainties. The IMC filter is conventionally used to en-
hance robustness properties by victimizing closed loop re- Remark4: For IMC structure the complementary sensitivity
sponse and making the compensator implementable (prop- function and the uncertainty can be computed as follows:
er). Moreover, it accounts for online adaptation of the con-
qp qp (7)
trol system by adjusting the filter time constant. In this pa- (s)
1 q( p ~
p ) 1 qp(1 1/ D )
per we extend this approach by using a second IMC filter
which assures the closed loop to be strictly proper and has a
lm ( s ) D 1 (8)
smooth response by compensating the singular plant im-
pulsive behavior. The singular internal model control filter Therefore, in this case condition (6) cannot be satisfied.
or SIMC filter is designed to yield a continuous smooth re- Thus we need to modify the IMC structure or algorithm in
sponse and a robust IMC design for singular systems. In order to gain a more tractable uncertainty profile. In the
fact by using a parallel strictly proper model in IMC, the following section the SIMC filter is introduced and the pro-
uncertainty will become unbounded and the robust control posed method is studied.
will not be feasible any more. Therefore the SIMC filter has
another role of bounding the uncertainty profile and mak- THE SIMC FILTER
ing the robust control problem feasible. The disk-type un- The idea of augmenting the IMC compensator by an IMC
certainty profile is usually assumed in robust control filter can be extended to singular systems in a different
schemes, is described by the following relation. manner. According to the previous discussions one way to
p ( j ) ~
p ( j ) overcome the obstacles in IMC of singular systems is to
lm ( ) l m ( ) (4)
~p ( j ) augment the compensator by an additional IMC filter, we
call it SIMC. This filter have the same structure as the con-
This uncertainty description allows us to incorporate sever- ventional IMC filter for step reference signals, and there-
al singular systems in the design while the state space un- fore, the IMC problem of singular systems consists of find-
IJSER 2011
234

ISSN 2229-5518
ing two time constants; One for the conventional IMC qp

( s) qp q~pD q~f1 ~
pDf 2 q~~
p f1
which adjusts the closed loop performance, robustness and 1 qf 2 ( p p )
noise amplification; and one for the feasibility of robust
control and impulse elimination of the singular plant. This Therefore condition (6) can be states as (12).
is expectable for a singular system to require more parame-
ters to be controlled, because a singular system is a general Remark7: The above lemma states an essential character
form of linear systems and cannot be treated by the same of SIMC, the SIMC filter caused the uncertainty to remain in
existing methods in standard systems. One advantage of a disk shaped region and the robust stability criterion is
SIMC is to solve the problem by introducing an additional then applicable to the problem. If one studies condition (6)
filter without any need of complicated design procedures. with and without SIMC filter, it can be seen that SIMC filter
Define the SIMC filter as a low pass filter: imposes a bound on the uncertainty. Also choosing D as the
1 inverse of SIMC filter, the uncertainty profile remains un-
f 2 ( s) (9)
(s 1) m changed and the uncertainty caused by singular system will
be zero as can be seen from (11).
Lemma6: Define SIMC filter as stated in (9), therefore the When there are some model inaccuracies or distur-
closed loop system is strictly proper, if and only if: bances, condition (12) cannot be met easily because a specif-
m (10) ic performance index is expected in the control objectives.
In which, parameter denotes the relative degree of the In these situations a natural compromise exists and the vic-
plant. timization of performance is inevitable. Note that one can
Proof: Using (9-10) as the SIMC filter, the relative degree set the IMC filter to zero in order to satisfy (12) but this
of plant/compensator becomes strictly proper. Therefore, means open loop control of the system and therefore losing
by means of lemma2, the closed loop system is strictly performance. The uncertainty bound generally increases
proper. with frequency. A natural routine for making the controller
Remark5: There is no need to introduce pole zero cancel- robust is to design a nominal H2-optimal controller accord-
lation issues because SIMC filter cancels minimum phase ing to performance specifications and then increasing the
zeros of the plant at most. filter time constant to meet the desired robustness proper-
Lemma7: Together with SIMC filter the singular plant is ties.
capable of being robustly controlled, if (6) can be satisfied. Theorem1: Assuming D f 21 then there is an IMC filter
Proof: The new uncertainty profile have the following such that the closed loop system is robustly stable, and fur-
shape: thermore, the system exhibits robust performance at the
p zero frequency, if and only if:
pf 2
lm ( s ) D Df 1 (11)
2 lm (0) 1 (13)
p
D Proof: The IMC filter should satisfy (12) for robust stability,
Now it is easy to choose SIMC filter such that the uncer- because of the structure selected for IMC filter, the maxi-
tainty profile is bounded. mum value for the filter is unity and it occurs at zero fre-
quency. Therefore, for the nominal plant (12) can be satis-
Remark 6: Note that the real uncertainty profile between fied only if the uncertainty upper bound is smaller than
actual plant and assumed singular model is unchanged. unity, and therefore, the necessary condition for the IMC
SIMC manipulates only the mismatch between singular filter to exist is (13). The sufficiency is obvious.
model and the implemented parallel strictly proper model Remark8: Note that theorem 1 is an extension of the existing
of IMC. Also note that l represents the uncertainty caused
m result in standard systems. Although the SIMC filter does
by singular system while l m is the actual uncertainty. not appear explicitly in the theorem, it has an essential role
in the derivation of the theorem as well as lemmas. In other
Lemma8: The closed loop system with SIMC structure
words, introducing the SIMC makes it possible to apply the
characterized by equations (9-11) is robustly stable, if and
existing frame for robust control to the singular systems.
only if:
1 Remark9: Theorem1 just considers the solvability of (12). In
f1 (12)
~ ~
pq lm other words it studies the existence of an appropriate IMC
filter which solves the robust control problem. In order to
Proof: The complementary sensitivity function can be find such an IMC filter one should increase the time con-
stated as: stant and check the robust stability criterion until it is satis-
fied.
IJSER 2011
235

ISSN 2229-5518
Remark10: It should be noticed that there exists no con- Corollary2: For a strictly proper plant/compensator, (15)
straint on the SIMC filter time constant and any positive does not occur because:
time constant can be chosen. However when smoothness of
the response is also a requirement, large time constant for cp( s ) a1s 1 a 2 s 2 .... (16)
SIMC filter is required, and when a fast response is desired, As a result for a strictly proper compensator/plant combi-
it is better to choose the time constant as small as possible. nation the regularity issue is not of concern. This corollary
Note that if the SIMC filter time constant is larger than that depicts the fact that why the regularity control objective is
of IMC filter and the plant dominant time constant, it will introduced only for singular systems and not for standard
determine the closed loop time constant. In fact the closed strictly proper ones.
loop time constant is the largest time constant among the
plant, IMC filter and SIMC filter time constants. Because of In the following theorem we may introduce the interesting
robustness considerations SIMC filter time constant may be characteristics of the proposed algorithm.
smaller than IMC filter time constant, and does not restrict Theorem2: The closed loop system with an appropriate
the closed loop performance. It is not possible to decrease IMC filter designed according to (12) is robustly strictly
SIMC filter time constant as much as desired, since input proper and robustly regular against all uncertainties de-
noises may be amplified. scribed by (4).
Remark11: Note that (13) means that steady state gains for Proof: Note that from theorem1, the closed loop robust sta-
the plant and model should be of the same sign. A little bility and zero frequency performance are assured. The
mismatch between plant and model steady state gain may family of plants described by (4) all have a singularity index
cause instability if their sign were different. This a common smaller than or equal to that of nominal plant. This can be
drawback of robust control systems for plants with zeros shown as follows; assume that there is a plant in the family
near the origin. By a slight change of the zero location the (4) that has a larger singularity index than the nominal
closed loop may become unstable if the zero is near the ori- plant. Then uncertainty profile can be written as:
gin.
p ~
p
Lemma9: Irregularity of closed loop occurs, if and only if: lm ~
p
cp 1 for all s
From the above assumption uncertainty will increase by
Proof: From the definition of regularity, a singular system is frequency because it has an improper transfer function.
irregular if and only if: Therefore, (4) cannot be satisfied as the uncertainty is un-
bounded. Moreover, for any plant being in family (4) the
sE A 0 (14) relative degree of SIMC filter is greater than or equal to the
plant singularity index and thus the closed loop system is
In the frequency domain context of output feedback control
robustly strictly proper according to lemma2. Also note that
systems the above determinant is the characteristic poly-
nomial of system or the denominator of complementary regularity of the plant is guaranteed by lemma8 because of
sensitivity function. Write the closed loop transfer function strict properness of plant/compensator combination.
as:
The following design procedure can be followed for robust
N internal model control of a singular plant.
cp N
(s ) M Design Procedure:
1 cp 1 N N M
M 1. Choose the polynomial D and set f 2 as its inverse. The
According to (14) and (15) the closed loop system is irregu- polynomial time constant should be smaller than the domi-
lar, if and only if: nant time constant of plant. According to nominal singular
plant choose m such that the strictly properness of closed
N M loop is guaranteed.
This can be rewritten as:
2. For the nominal plant check the feasibility of robust con-
cp 1 for all s (15) trolhaving uncertainty profile as (4) according to (12), if
satisfies design IMC filter for a good performane in nominal
The last equality also means an unsolvable algebraic loop in case.
the simulation.
3. Redesign SIMC filter for having better performance if
required.
IJSER 2011
236

ISSN 2229-5518
SIMULATION RESULTS
Simulating an improper system is not possible with the
existing numerical methods, since simulation needs future
data for computing the present state vector. This is why
many Papers in the field of singular systems do not include
any simulation examples or just simulate causal singular
systems. However, if the closed loop system is proper, any
simulating software can easily implement the closed loop
system regardless of the inner unsolvable loops, which
form singular systems in the inner parts of the closed loop
system. In this paper, some illustrative simple examples are
chosen in order to show the effectiveness of the proposed
algorithm.
Example1: Consider the nonlinear system described by the

following equations:
x1 6 x1 2 x2 u x12
x2 u 0
y x1 2x2 (17)
The output equation is the simplest form of output equa-

tion. Nonlinear output may occur in a singular system;
however they are essentially treated in a similar manner.
The algebraic part of a singular system denotes its limita-
tions for having arbitrary initial conditions. The system de-
scribed by (17) can be modeled by a standard state space
system, too. For a nominal input of u=9, the equilibrium
point will be:
x1 x 2 0 x1* 3, x2* 9 (18)

In case of nonlinear term in (17), the nonlinearity can be
considered as an uncertainty, and not included in the linear
model representation.
The process model may be considered as a bi-proper trans-

fer function.
~ 2 s 13 2 s 13
p(s) p( s )
( s 6)(0.1s 1) (s 6)
IJSER 2011
237
International Journal of Scientific & Engineering Research Volume 2, Issue 4, April-2011 7
ISSN 2229-5518
The compensator, IMC and SIMC filters may be chosen as: 8
7
1 1
f 2 ( s) 6
(s 1) D
Step response
5
(0.1s 1)(s 6)
q~ ( s)
4
2 s 13 3
1 2
f1 ( s) 1
(0.5s 1) 0
Closed loop system responses to different inputs with dif- -1

0 10 20 30 40 50 60 70
ferent initial conditions are shown in figures 1 to 3. Before Time (sec)
change in the set point, initial condition is vanished and Figure 3: Set point response with initial condition
then the set point signal is tracked without any off set. The
closed loop system is stable as the phase portrait shows and
0.4
as it is strictly proper, a smooth response is attained. Closed
loop system is able to follow any piecewise constant refer- 0.2
ence signal and reject constant disturbances. The main ad- 0
vantage of the presented control structure is to make closed

-0.2
loop, impulse free, even strictly proper, in a robust manner,
x2
i.e. regardless of uncertainty in the plant model. -0.4
-0.6
6 -0.8
5 -1
-5 -4 -3 -2 -1 0 1 2 3 4 5
X1
4
Figure 4: phase portrait of closed loop near the origin
Initial condition response
2 This example shows that the closed loop system is strictly

1 proper regardless of any model uncertainties while bounded and
0
also it provides an example of robust stability and zero frequen-
cy performance design.
-1
-2
Example2: Consider a group of linear singular systems as
0 10 20 30 40 50 60 70
Time (sec) described by the following set of transfer function.
Figure 1: System response to initial condition
p ( s ) s 2 s 1
12 p1 ( s) s 2 3s 1
10 Disturbance rejected p2 ( s) s 2 4s 1
Step response with disturbance
8
p3 ( s ) s 2 1
Disturbance occured
6
The nominal plant (model) is assumed to be:
~
p( s) s 2 2s 1
4
2 The parallel model and compensator are:
~ s 2 2s 1
0 p ( s)
0 10 20 30
Time (sec)
40 50 60
(0.1s 1) 3
Figure 2: Set point and disturbance response
(0.1s 1) 3
q ( s ) f1 ( s)
s 2 2s 1
It is aimed to design a single robust controller for all plants
described above. The uncertainty norm is bounded for all of
the models described in (19), however, its infinity norm is
near the unity for case p4. Following figures depict closed
IJSER 2011
238
International Journal of Scientific & Engineering Research Volume 2, Issue 4, April-2011 8
ISSN 2229-5518
loop behavior in tracking step set point. Set point tracking is CONCLUSION
almost perfect even in presence of uncertainties. This charac-
In this paper a new, effective and simple control scheme is
teristic also exists in conventional unit feedback control, e.g.
proposed for robust internal model control of singular linear
PID controllers. However state space methods like [5] dont
systems. The method has many advantages over the exist-
include this feature. In contrast to the aggressive nature of
ing, state space methods including robust strict properness
singular systems, closed loop response is smooth enough to
of the closed loop, avoiding algebraic loops, robust tracking
ensure preventing any damage to the instruments.
1.4
of specific signals and the ability to robustly stabilize a larg-
er group of singular systems comparing with other methods.
1.2
Two simulation examples are included to depict the algo-
1 rithm performance.
Step Response
0.8
References
0.6
0.4
[1] Mertzios, B.G. and F. Lewis, Fundamental matrix of discrete singular
0.2
systems, Journal of circuits, systems and signal processing, vol.8, NO.3,
1989.
0
0 20 40 60 80 100 120 140 160 180 200
[2] Dai, L. Singular control systems, Springer verlag, 1989.
Time (sec)
[3] Campbell, S.L, R. Nikoukhah and B.C. Levy, Kalman filtering for
general discrete time linear systems, IEEE transactions on automatic
Figure 5: step response for p1
control, vol.44, NO.10, pp. 1829-1839,1999.
1.4
[4] Luenberger, D. Dynamic equations in descriptor form,IEEE
1.2 transactions on automatc control, vol.AC.22, NO.3, June, 1977
[5] Syrmos, V.L. and F.L. Lewis, Robust eigen value assignment in
1
generalized systemsProceedings of the 30th conference on decision and
control, England, 1991.
Step Response
0.8
[6] Fang, C.H., L.Lee and F.R. Chang, Robust control analysis and design
0.6 for discrete time singular systems, Automatica, Vol.30, NO.11, pp.1741-
1750, 1994.
0.4
[7] Xu, S. , C. Yang, Y. Neu and J. Lam, Robust stabilization for uncertain
discrete singular systems, Automatica, Vol.37, pp.769-774, 2001.
0.2
[8] Mukundan, R. and W. Dayawansa, Feedback control of singular
0 systems- proportional and derivative feedback of the state, International
0 20 40 60 80 100 120 140 160 180
Time (sec) journal of systems science, vol.14, NO.6, pp.615-632, 1983.
[9] Chu, D.L., H.C. Chan and D.W. Ho, Regularization of singular systems
Figure 6: Step response for p2 by derivative and proportional output feedback, Siam journal of matrix
1.4
nalysis and applications, vol.19, NO.1, pp.21-38, January, 1998.
1.2
[10] Xu, S. and J. Lam, Robust stability and stabilization of discrete singular
systems: An equivalent characterization, IEEE transaction on automatic
1 control, vol.49, NO.4, April 2004.
[11] Hou, M. Controllability and elimination of impulsive modes in
Step Response
0.8
descriptor systems, IEEE transactions on automatic control, vol.49,
NO.10, 2004.
0.6
[12] Morari, M. and E. Zafiriou, Robust process control, Prentice Hall,
0.4 1989.
0.2
0
0 50 100 150 200 250
Time (sec)
Figure 7: Step response for p3
In the last case the oscillating behavior of plant is not in-

cluded in the model and therefore closed loop response is
not satisfactory. Note that while steady state gains of plant
and model have the same sign, the closed loop is robustly
stable. And while the uncertainty is bounded it is robustly
strictly proper.
IJSER 2011
239

ISSN 2229-5518
An Integrated Partial Backlogging Inventory

Model having Weibull Demand and Variable
Deterioration rate with the Effect of Trade Credit
P.K. Tripathy, S. Pradhan
Abstract - Demand considered in most of the classical inventory models is constant, while in most of the practical cases the
demand changes with time. In this article, an inventory model is developed with time dependent two parameter weibull demand
rate whose deterioration rate increases with time. Each cycle has shortages, which have been partially backlogged to suit
present day competition in the market. Also the effect of permissible delay is also incorporated in this study. The total cost
consists of ordering cost, inventory holding cost, shortage / backordering cost, lost sale cost and deterioration cost are
formulated as an optimal control problem using trade credit policy. Optimal solution for the model is derived and the trade credit
on the optimal replenishment policy are studied with the help of numerical examples.
Index Terms- Inventory, Shortages, partial backlogging, weibull demand, trade credit , variable deterioration, replenishment

1. INTRODUCTION In some real life situation there is a part of demand which
can not be satisfied from the current inventory, leaving the
I n the classical inventory economic order quantity system in stock out. In these systems two situations are
(EOQ) model, it was tacitly assumed that the customer mainly considered, all customers wait until the arrival of
must pay for the items as soon as the items are received. the next order (Complete back order case) or all
However, in practices or when the economy turns sour, customers leave the system (lost sales case). However,
the supplier frequently offers its customers a permissible in practical, some customers are able to wait for the next
delay in payments to attract new customer. In todays order to satisfy their demands during the stock out period,
business transaction, it is frequently observed that a while others do not wish to or can not wait and they have
customer is allowed some grace period before settling the to fill their demands from other sources. This situation is
accounts with the supplier or the producer. The customer modeled by consideration of partial backordering in the
does not have to pay any interest during this fixed period formulation of the mathematical model. Wee (1999)
but if the payment gets delayed beyond the period developes a determinatic inventory model with quantity
interest will be charged by the supplier. This arrangement discount, pricing and partial backlogging when the
comes out to be very advantageous to the customer as product in stock deteriorates with time according a weibull
he may delay the payment till the end of the permissible distribution. Teng (2002) presents on EOQ model under
delay period. During the period, he may sell the goods, the conditions of permissible delay. Chen et al. (2003)
accumulate revenues on the sales and earn interest on establish an inventory model having weibull deterioration
that revenue. Thus it makes economic sense for the and time varying demand. Wu et al. (2003) considered an
customer to delay the payment of the replenishment inventory model where deteriorating rate and demand
account upto the last day of the settlement period allowed rate are follows the Chens model (2003) where shortages
by the supplier on the producer. Similarly for supplier, it are permitted. Papachristos and Skouri (2003) present a
helps to attract new customer as it can be considered production inventory model with production rate, product
some sort of loan. Furthermore, it helps in the bulk sale of demand rate and deteriorating rate, all considered as
goods and the existence of credit period serves to reduce functions of the time. Their model follows shortages and
the cost of holding stock to the user, because it reduces the partial rate is a hyperbolic function of the time up to
the amount of capital invested in stock for the duration of the order point. They propose an algorithm for finding the
the credit period. So, the concept of permissible delay in solution of the problem. Abad (2003) considers the
payments beneficial both for supplier and customer. problem of determining the optimal price and lot size for
reseller in which unsatisfied demand is partially
----------------------------------------------- backordered. There are several interesting papers related
1. P.G. Department of Statistics, Utkal Universitym to partial backlogging and trade credits viz. park (1982),
Bhubaneswar-751004, India, [email protected] Jamal et al. (1997) , Lin et al. (2000), Dye et al. (2007)
2. Dept. of Mathematics, Orissa Engineering College,
Bhubaneswar-751007, India, [email protected] and their references.
With this motivation, in this paper an attempt is

made to formulate an inventory model in time
dependence of demand follows a two parameters weibull
type with time variable deterioration where unsatisfied
IJSER 2011
240

ISSN 2229-5518
demand is partially backlogged and delay payment is (Fig-1)
allowed. With the help of this model, supplier can easily
extract its order quantity, cycle period for shortages as
well as payment time to reduce the total inventory cost 3. MODEL FORMULATION:
In this model two parameter weibull type of
2. NOTATION AND ASSUMPTIONS: demand is considered with variable rate of deterioration.
To develop the proposed model, the following Depletion of the inventory occurs due to demand (supply)
notations and assumptions are used in this article. as well as due to inventory i.e. during the period (0,T) and
shortages occurs during period (T1, T).
2.1 Notations The differential equation of inventory level I (t) w.r.t. time
I(t) : Inventory level at time t is given by
Q (t) : Order quantity at time t = 0 dI t
vt I t = t 1 (1)
A, h,C1, dt
C2, C3 : Ordering cost, holding cost, deteriorating cost, With boundary condition I (0) = Q and I (T1) = 0
shortage cost for backlogged items and the unit The solution of equation (1) is
cost of lost sales respectively. All of the cost v 2
parameters are positive constants. I t
2

t T1 2

Ie : Interest which can be earned (2)
Ic : Interest charges which invested in inventory, vt 2TI
Ic Ie

T1 t
2
,0 t T
2
M : Permissible delay in setting the amounts, 0<M< T v T1
T : Length of the replenishment cycle and Q T1 (3)
2
T1 : Time when inventory level comes down to zero, 0
< T1< T Similarly the differential equation for shortages is given by
TIC1 : Total inventory cost when M < T1 dI t t 1
(4)
TIC2 : Total inventory cost when T1, < M < T dt 1 T t
The solution of equation (4) for shortage is
2.2 Assumptions: t 1
1. The inventory system involves only one item I(t) = t T 1 (5)
1
2. Replenishment occurs instantaneously on ordering
i.e. lead time is zero and it takes place at an The deteriorating cost during one cycle is
T1
infinite rate. C vT1 2
DC = C1 Q t 1dt (6)
3. The demand rate of any time t is t 1 two 0 2
parameter weibull type, where 0 1, 0 The holding cost for carrying inventory over the period
are called scale and shape parameter respectively. (0,T) is given by
4. Q (t) = vt is the variable deterioration rate, where 0 T1
T1 1 vT1 3 9
< v <<1. HC = h I t dt h (7)
5. Shortages are allowed and unsatisfied demand is O 1 6 3
1 The total amount of shortage cost during the period (T1,
backlogged at the rate of T) is given by
1 T t
T
The backlogging parameter is a positive SC = C 2 I t dt
constant, where T1 t T . T1
T 1 1
1 1 T T 1
C2 (8)
2
T
2
T
1 2
1

The amount of lost cost during the period (0,T) is given by
T
t 1
LC C 3 1
T1 1 T t
T 1 T1
= C3 T1 T (9)
1 1
IJSER 2011
241

ISSN 2229-5518
The solutions for the optimal values of T1 and T
Case I 0 < M < T1 can be found by solving the following equations
In this situation since the length of period with simultaneously.
positive stock is larger than the credit period, the buyer T IC1 TIC1
0 and 0 (13)
can use the sale revenue to earn interest at an annual T T1
rate Ie in (0,T1) Provided they satisfy the conditions
The interest earned IE1 , is
M
2TIC 1 2TIC1
PI M 1 0 and 0 (14)
IE 1 PI e t dt e (10) T 2 T1
2
O 1
Case II: T > M >T1
In this situation interest charged IC2=0 and the
interest earned per time unit is
T
IE 2 PI e t dt T M T
0
2T 1
PI e MT (15)
1
(Fig-2)
However, beyond the credit period, the unsold

stock is supposed to be financed with an annual rate Ic
and the interest charged IC1 , is
T1
IC 1 PI c I t dt
M
T 1 v 9 T 3
1 1

1 6 3
PIc 3
v M 2
M .T
2 3 1

M 1 v T1 M 3 (Fig-3)
T1 M (11)
1 6
Total inventory cost per unit time is given by Then the total inventory cost per unit time is
1 given by
TIC1 A HC SC LC DC IC1 IE1 1
T TIC 2 A HC SC LC DC IC 2 IE 2
T
1 T1 1 v 9T1 3 T 1 1
A h
6 3
c2
T1 T 1 I T1 1 v 9T1 3
T 1 1 A h
T 1 6 3

T 2 T1 2 T 1
T1 T
T1
+ C3 T 1 1
T 2 T1 2
1 2 1 1 C2
T1 T 1
1 1 2
v T1 2 T1 1 v 9 T1 3
C1 PIC T 1 T1 vT1 2
2 1 6 3 C3 T1 T C1

1 1 2
v M 3 2 M 1 v M 3T1
MT1 T1 M 2T 1
2 3 1 6 PI e M T (16)
1 1
PI e M
(12)
1 The solutions for the optimal values of T1 and T
can be found by solving the following equations
simultaneously.
IJSER 2011
242

ISSN 2229-5518
2TIC 2 2TI C 2 of the model analysis by obtaining the optimal inventory

2
0 and 0 (17) length ,optimal economics order quantity, shortage time
T T12
period and also calculated the total variable cost. From
the numerical examples it is found that for high values of
Provided they satisfy the conditions the parameter , zero terminal inventory level provides
2TIC 2 2TIC 2 higher unit of the total variable cost. Thus for simplicity
0 and 0 (18)
T 2
T12 regarding the applicability of the model it is not preferable
to adopt zero-terminal inventory level policy.
Numerical Examples:
The proposed model can be extended in
Case I : A=50, h=2, c1=2, C2=0.8, C3=2.0, =0.0002, numerous ways. For example we may consider the effect
v=0.001, =0.5, p=0.5, Ic= 0.2, Ie=0.1, m =0.05 of inflation and time value of money by taking discount
rate.
T1 T Q TIC1
0.8 1.90535191 106.840667 0.000335 0.718680
1.0 2.2908332 77.122890 0.000459 1.130076 APPENDIX - I
1.5 2.917128 40.368203 0.001 2.662553
2.0 3.01722969 24.922125 0.001829 4.927522 The demand rate function D(t) =t -1, , > 0 follows
2.5 2.77056753 17.154589 0.002566 7.866730 two parameters weibull distribution in time t.
3.0 2.4129991 12.736923 0.00282 11.374538
3.5 2.07280312 9.991002 0.002571 15.332234 Thus functional form represents increasing, decreasing
4.0 1.78842979 8.166399 0.00205 19.628761 and constant demand for different values of the
4.5 1.56061254 6.889604 0.001485 24.170034
5.0 1.379736015 5.958621 0.001001 28.880925
parameter
Case II : We have
A=50, h-2, c1=2, C2=0.8, C3=2.0, =0.0002, v=0.001, dD(t )
( 1)t 2 (19)
dt
=0.5, p=0.5, Ie=0.1, m =0.5 d 2 D(t )
and ( 1)( 2)t 3 (20)
T1 T Q TIC2 dt 2
0.8 0.4998753 1.257346 0.000115 39.766547
1.0 0.41972384 1.066637 0.000084 46.876526 D (t ) d 2 D( t )
1.5 0.398675 1.051521 0.00005 46.962943 i) for 0<<1, 0 and 0 the demand rate
2.0 0.37985845 1.047747 0.000029 47.721766
dt dt
2.5 0.359786885 1.007243 0.000016 49.640806 decreased with time at an increasing rate.
3.0 0.34985674 0.990081 0.000009 50.501263
3.5 0.29985764 0.85216 0.000003 58.674548 ii) =1, the demand rate becomes steady over time.
4.0 0.2789685 0.797059 0.000001 62.730769 d D( t ) d 2 D( t )
4.5 0.2756759 0.791841 0.000001 63.144167 iii) 1< 2, 0, 0, then the demand
dt dt 2
5.0 0.2679865 0.772816 00000 64.698610
increases with time at decreasing rate.
4. CONCLUDING REMARKS: Case (i) is applicable to the seasonal products towards

the end of the seasons.
In this paper, an EOQ inventory model is Case (ii) is applied to the consumer goods long
considered for determining the optimal replenishment established in the market by their brand name.
schedule under variable deterioration with two parameter Case (iii) is appropriate for some fashionable product
weibull type demand where shortages are partially back lunched in the market
logged. Also the proposed model in- cooperates other
d D( t ) d 2 D( t )
realistic phenomenon and practical features such as trade If >2 then 0 and 0, which implies
credit period. The credit policy in payment has become a dt dt 2
very powerful tool to attract new customers and a good that the demand will go on increasing with time at an
incentive policy for the buyers. In keeping with this reality, increasing rate. Thus there will be an accelerated growth
these factors are incorporated into the present model. in demand, which is not very realistic.
The model is very useful in retail business. In real market
situations demand rate may be increased, decreased or
constant. Keeping these consideration, we have assumed REFERENCE
here that the time dependence of demand follows the two [1] A.M.Jamal,B.R.Sarker and S.Wang ( 1997).An
parameters weibull distributions where depending upon ordering policy for deteriorating items with allowable
the values of the parameter (1,1 and 2 ),the degree shortage and permissible delay in payment, Journal
of non-linearity varies. (see Appendix I) Numerical of the Operational Research Society,48,826-833.
examples are presented to justify the claim of each case
IJSER 2011
243

ISSN 2229-5518
[2] Jimm-Tsair Teng and Liang-yuh ouyang(2005).An
EOQ model for Deteriorating items with Power-Form
Stock-Dependent demand ,Information and
Management Sciences, 16,1-16.
[3] S.K.Goyal and B.C.Giri(2001).Recent trends in
modeling of deteriorating inventory, European
Journal of Operational Research,134,1-16.
[4] R.B Covert and G.S.Philip(1973).An EOQ model with
weibull distribution deterioration, AIIE
Transaction,5,323-326.
[5] T.K.Datta and K.Paul(2001).An inventory system with
stock- dependent, price sensitive demand rate,
Production Planning and Control, 12, 13-20.
[6] I. T. Teng(2002). An Economic order quantity under
conditions of permissible delay in payments, Journal
of the Operational Research Society. 53, 915-918.
[7] J. M. Chen and S. C. Lin(2003). Optimal
replenishment scheduling for inventory items with
weibull distributed deterioration and time varying
demand, Journal of Information and Optimization
Sciences, 24 ,1-21.
[8] J. W. Wu and W. C. Lee(2003). An EOQ inventory
model for items with weibull distributed deterioration,
shortages and time varying demand. Information
and optimization Science 24, 103-122 .
[9] S. Papachristos and K. Skouri(2003). An inventory
model with deteriorating items, quantity discount ,
pricing and time dependent partial backlogging,
International Journal of Production Economics, 83,
247-256.
[10] C. Y. Dye , L. Y. Ouyang and T. P. Hsieh(2007).
Deterministic inventory model for deteriorating items
with capacity constraint and time proportional
backlogging rate, European Journal of Operational
Research, 178, 789-807.
[11] K.S Wu, L.Y ouyang and C.T Yang(2006).An optimal
replenishment policy for non-instantaneous
deteriorating items stock dependent demand and
partial back logging , International Journal of
Production Economics,101,369-384.
[12] P.L.Abad(2003).Optimal pricing and lot-sizing under
conditions of perishability, finite production and partial
back ordering and lost sale, European Journal of
Operational Research, 144, 677-685.
[13] S. K. Park (1982) Inventory model with partial
backorders. International Journal of Systems
Science, 13, 1313-1317.
[14] H. M. Wee (1999). Deteriorating inventory model with
quantity discount, pricing and partial backordering.
International Journal of Production Economics, 59,
511-518 .
[15] C. Lin, B. Tan and W. C. Lee(2000). An EOQ model
for deteriorating items with time varying demand and
shortages, International Journal of systems Science,
31, 391-400.
IJSER 2011
244

ISSN 2229-5518
Survey of Current and Future Trends in

Security in Wireless Networks
Vikas Solomon Abel
Abstract Security has always been a key issue with all the wireless networks since they have no physical boundaries. Many existing and
evolving threats which must be considered to ensure the countermeasures are able to meet the security requirem ents of the environments for
which it is expected to be deployed.
Index Terms Bluetooth, Wi-Fi, WiMax, Wireless Sensors, WLAN

connection to the wired network. For wireless devices in a
1 Wireless Technologies WLAN to communicate with each other, they must all be
configured with the same SSID. Wireless devices have a
default SSID that is set at the factory. Some wireless devices
W ireless Broadband is a fairly new technology that
provides high-speed wireless internet and data
refer to the SSID as network name.
network access over a wide area. The three most important

Wi-Fi or Wireless Fidelity is to connect to the internet at
broadband wireless technologies are IEEE 802.11, IEEE
high speeds. Wi-Fi enabled devices operates in unlicensed
802.16, and Wireless Mesh Network (WMN). In Wireless
spectrum. They use radio technologies within the range of
Broadband Networks Privacy and Integrity of data are key
an access point for data communication which are based
considerations for both broadband networks.
on the IEEE 802.11 standards. The benefits of using Wi-Fi in
Authentication and Security to prevent unwanted access to
the last mile are that, the client device is so cheap due to
critical data or services are necessary for the effective
large volume of production [3].
operation of any broadband network. The motivation
behind Wireless Broadband technology was the support of
Sensor networks often have one or more points of
mobile clients within a certain range and to provide
centralized control called base stations. A base station is
wireless network infrastructure in such places where wired
typically a gateway to another network, a powerful data
networks are not feasible or possible.
processing or storage center, or an access point for human
interface. They can be used as a nexus to disseminate
WLAN which is IEEE 802.11 technology consist of two
control information into the network or extract data from it.
important components, i.e. nodes and an Access Point
They may also be referred to as sinks. Base stations are
(AP). The AP provides wireless connectivity to nodes as
typically more powerful than sensor nodes. Sensors are
well as function as a bridge between the nodes and wired
constrained to use lower-power, lower-bandwidth, shorter-
network infrastructure. Probe request frames are used by
range radios. Sensor nodes form a multihop wireless
the clients to discover a wireless network. If a wireless
network to allow sensors to communicate to the nearest
network exist then the AP respond with Probe response
base station.
frame. The AP (Access Point) which provides the strongest
signal is selected. If the AP is overloaded then the nodes
Aggregation points help in reducing the total number of
connect with another nearby AP although the signal
messages sent thus saving energy, sensor readings from
strength is relatively weak [1]. Typical applications of
multiple nodes may be processed at one of many possible
WLAN are campus and organizational networking; back
aggregation points. An aggregation point may collects
haul for public safety and industrial control networks [2].
sensor readings from surrounding nodes and forwards a
WLAN may be an Infrastructure WLAN where client
single message representing an aggregate of the values. It is
devices communicate wirelessly to a wired Local Area
also possible that every node in the network functions as an
Network (LAN) such as Ethernet, via Access Points or an
aggregation point, delaying transmission of an outgoing
Ad-Hoc WLAN in which devices or stations communicate
message until a sufficient number of incoming messages
directly with each other, without using Access Points or a
IJSER 2011
245

ISSN 2229-5518
have been received and aggregated. Power management in

sensor networks is vital [4].
WMN is an emerging broadband wireless technology

which consists of mesh routers and mesh nodes. The mesh
routers form the backbone and provide connectivity to
mesh nodes. With some of the mesh routers in WMN the
functionality of gateways is provided for internet
BS
connectivity to mesh nodes. The WMN nodes act both as a
client and as a relay router for forwarding traffic for other
nodes. In most of the cases, the WMN form partial mesh
topology, which means that there are many routes to every MS
node, so if any route is not working then the data traffic can
be sent through other route [2]. RS
Bluetooth is a universal radio interface in the 2.45 GHz

frequency band that enables portable electronic devices to RS
connect and communicate wirelessly via short-range, ad MS MS
hoc networks. Each unit can simultaneously communicate BS Base Station
with up to seven other units per piconet. Moreover, each RS Relay Station MS Mobile Station
unit can simultaneously belong to several piconets [5].
Fig 1. Multi-hop Relay Network architectures
Source : [23]
Bluetooth uses a radio technology called frequency-
hopping spread spectrum. A Bluetooth device will use 2 General Attacks on Wireless Networks
many different and randomly chosen frequencies every General, attacks on wireless networks may be divided into
second to minimize the probability of other devices using four basic categories: passive attacks, active attacks, man-
the same frequency and to minimize interference time. For in-the middle attacks, and jamming attacks.
battery operated devices Bluetooth is ideal. It does not rely
on the user input since it can automatically detect and A passive attack occurs when someone listens to or
communicate with other Bluetooth devices. The frequency eavesdrops on network traffic. A passive attack on a
of Bluetooth capable devices ranges from 2.402 GHz to as wireless network may not necessarily be malicious in
high as 2.480 GHz, a frequency range specifically reserved nature. They are very difficult to detect.
by international agreement for ISM or medical, industrial
and scientific devices. In passive attacks the attacker is not concerned with the
routing protocol. It only eavesdrops on the routing traffic
WiMAX, an acronym for Worldwide Interoperability for and endeavours to get valuable information like node
Microwave Access, is a telecommunications technology hierarchy and network topology from it. For example, if a
that provides fixed and fully mobile internet access. The route to a particular node is requested more than other
IEEE 802.16 standard forms the basis of . Clarification of the nodes, the attacker might predict that the node is important
formal names are as follow: for the operation of the network, and putting it out could
bring down the entire network. Even when it is not
802.16-2004 is also known as 802.16d, which refers to the possible to isolate the precise position of a node, one may
working party that has developed that standard. Since it be able to determine information about the network
has no support for mobility it is sometimes referred to as topology by analysing the contents of routing packets. This
"Fixed WiMAX". 802.16e-2005, is also known as 802.16e, is attack is very difficult to detect and prevent [6].
an amendment to 802.16-2004. It introduced support for In active attacks, the aggressor node has to spend some of
mobility, among other things and is therefore also known its energy in order to carry out the attack. Nodes that
as "Mobile WiMAX". 802.16j-2009, the Multihop Relay perform active attacks with the aim of disrupting other
specification for 802.16. nodes by causing network outage are considered to be
harmful, while nodes that perform passive attacks with the
IJSER 2011
246

ISSN 2229-5518
aim of saving battery life for their own communications are 3.5 Denial of Service Attack
considered to be selfish. The harmful or malicious nodes In a "denial-of-service" attack an attacker may try to
can change the routing information by modifying, prevent an authentic user from using a service. This may be
fabricating or impersonating nodes and thereby disrupting done by "flooding" a network, disrupt connections between
the correct functioning of a routing protocol [6]. two machines etc. The common method is to flood a
network with degenerate or faulty packets and hence
Once an attacker has gained enough information from the denying access to the legitimate traffic.
passive attack, the hacker can then launch an active attack
against the network. There are an large number of active 3.6 Physical Tampering
attacks that a hacker can launch against a wireless In this attack the physical device may be tampered and
network. Some examples of the attacks are Spoofing attacks, sensitive information can be extracted from it.
Unauthorized access, Flooding attacks etc. Spoofing occurs
when a malicious node misrepresents its identity by 4 General Link layer Attacks
altering its MAC or IP address in order to mislead the 4.1 Unencrypted Management Communication
information about the network topology. It can also create Almost all the IEEE 802.16 management messages are still
routing loops etc. sent unencrypted. The IEEE 802.16-2004 standard does not
provide any capability to encrypt management messages.
3 General Physical layer Attacks
3.1 Jamming Attacks on Wireless Networks 4.2 Masquerading threat
Jamming is a attack specific to wireless networks. Jamming By intercepting the management messages an attacker can
occurs when spurious RF frequencies interfere with the use the hardware address of another registered device.
operation of the wireless network. In some cases, the Once this is successful an attacker can turn a Base Station
jamming may be unintentional and is caused by the (BS) into a Rogue Base Station [7]. A rouge WiMAX base
presence of other devices such as cordless phones, that station pretends to be a valid base station, and then
operate in the same frequency as the wireless drops/eliminates all/some of the packets.
network. Intentional and malicious jamming occurs when
an attacker analyzes the spectrum being used by wireless 4.3 Threat due to Initial Network Entry
networks and then transmits a powerful signal to interfere In IEEE 802.16j-2009 no integrity protection for
with communication on the frequencies discovered. management messages can be made in case of multicast
transmissions and in case of initial network entry by a new
3.2 Scrambling candidate node [8].
Scrambling is a type of jamming attack for short intervals to
disorder targeted frames (mostly management messages)
5 General Network layer Attacks
which ultimately lead to network failure.
The main network-layer operations in MANETs are ad hoc
routing and data packet forwarding, which interact with
3.3 Water Torture Attack
each other and fulfill the functionality of delivering packets
In this attack the attacker pushes a Subscriber Station (SS)
from the source to the destination. Routing messages in Ad
to drain its battery or consume computing resources by
Hoc routing protocols exchange routing messages between
sending bogus frames [7].
nodes and maintain routing states at each node accordingly.
3.4 Man in the Middle Attack Data packets are forwarded by intermediate nodes along
In this attack an attacker intercepts the valid frames and an established route to the destination based on the routing
then intentionally resends the frames to the target system. states. Both routing and packet forwarding operations are
A wireless specific variation of man-in-the-middle attack vulnerable to malicious attacks, leading to various types of
is placing a rogue access point within range of wireless malfunction in the network layer. Network-layer
stations. If the attacker knows the SSID in use by the vulnerabilities generally fall into one of two categories:
network, he can gain information such as the key routing attacks and packet forwarding attacks, based on the
information, authentication requests and so on. It may target operation of the attacks [9].
involve spoofing an IP address, changing a MAC address
to emulate another host, or some other type of modification.
IJSER 2011
247

ISSN 2229-5518
In the context of DSR [11], the attacker may modify the 5.1 Blackhole Attack
source route listed in the RREQ or RREP packets by In networking, black holes refer to places in the network
deleting a node from the list, switching the order of nodes where incoming traffic is silently discarded (or "dropped"),
in the list, or appending a new node into the list [12]. In hiding from the source the information that the data did
AODV [10] are used, the attacker may advertise a route not reach its intended recipient. The black holes themselves
with a smaller distance metric than its actual distance to the are invisible, and can only be detected by monitoring the
destination, or advertise routing updates with a large lost traffic. An attacker creates forged packets to
sequence number and invalidate all the routing updates impersonate a valid mesh node and subsequently drop
from other nodes [13]. By attacking the routing protocols, packets [14].
the attackers can pull traffic toward certain destinations in
the nodes under their control, and cause the packets to be
forwarded along a route that is not the optimal route or N1
which may even be nonexistent. The attackers can also
create routing loops in the network [9].
D
S
Attacks may be launched in addition to routing attacks
such as in packet forwarding operations. These attacks do
not disrupt the routing protocol but poison the routing
states at each node. Instead, they cause the data packets to
be delivered in a way that is intentionally inconsistent with S - Source BH N2
N1 - Node 1
the routing states. The attacker may drop or modify the N2 - Node2
contents of the packets or duplicate the packets it has BH - Blackhole
D Destination
already forwarded. Attacker may also inject large amount
Fig 2. Blackhole Attack
of bogus packets into the network which wastes a Source : [22]
significant portion of the network resources and introduces
severe wireless channel contention and network The properties of Blackhole attack is that firstly, the
congestion in the MANET [9]. Blackhole node exploits the ad hoc routing protocol, such
as AODV, to advertise itself as having a valid route to a
Routing protocols for ad-hoc networks are based on the destination node, even though the route is false, with the
assumption that intermediate nodes do not maliciously intention of intercepting packets. Secondly, the packets are
change the protocol fields of messages passed between consumed by the Blackhole node. Thirdly, the Blackhole
nodes. This assumed trust permits malicious nodes to nodes can conduct coordinated attacks.
easily generate traffic subversion and denial of service (DoS)
attacks. Attacks using modification are generally targeted 5.2 Greyhole Attack
against the integrity of routing computations and so by Grey Hole is a node that can change from behaving
modifying routing information an attacker can cause correctly to behaving like a black hole so as to avoid
network traffic to be dropped, redirected to a different detection. Some researchers discussed and proposed a
destination, or take a longer route to the destination solution to a black hole attack by disabling the ability for
increasing communication delays. Forged routing packets intermediate nodes to reply to a Route Reply (RREP) only
may be sent to other nodes to divert traffic to the attacker allowing the destination to reply [15].
or to some other node. The intention is to create a black
hole by routing all packets to the attacker and then 5.3 Wormhole Attack
discarding it. As an extension to the black hole, an attacker In a wormhole attack, an attacker forwards packets
could build a grey hole, in which it intentionally drops through a out-of-band high quality link and replays those
some packets but not others, for example, forwarding packets at another location in the network [16].
routing packets but not data packets. A more subtle type of
modification attack is the creation of a tunnel (or wormhole)
in the network between two colluding malicious nodes
linked through a private network connection [6]. A brief
description of the attacks is as follows -
IJSER 2011
248

ISSN 2229-5518
The transport layer is responsible for managing end-to-end

N connections. Two possible attacks in this layer, flooding
N
and desynchronization as in case of sensor networks .
N
N
N 6.1 Flooding
N By repeatedly make new connection requests until the
resources required by each connection may be exhausted,
N may reach a maximum limit or may lead to memory
Fig 3. Wormhole Attack
exhaustion through flooding. In either case, further
Wormhole attacks depend on a node misrepresenting its legitimate requests will be ignored. One proposed solution
location. Hence, location based routing protocols have the to this problem is to require that each connecting client
potential to prevent wormhole attacks . demonstrate its commitment to the connection via the
solving of a puzzle. The idea is that a connecting client will
5.4 Flooding Attack not needlessly waste its resources creating unnecessary
In a flooding attack, the attacker targets the target node by connections. Since an attacker does not likely have infinite
flooding in order to congest the network and degrade its resources, it will be impossible for them to create new
performance [17]. connections fast enough to cause resource starvation on the
serving node. Although these puzzles do include
5.5 Rushing Attack processing overhead, this technique is still more desirable
In Multicast forwarding a single stream of data is sent to than excessive communication [20].
multiple receivers so as to save network bandwidth. A
rushing attacker exploits duplicate suppression mechanism 6.2 Desynchronization
by quickly forwarding route discovery packets and hence Disruption of an existing connection is refered to as
gaining access to the forwarding group. The goal of this Desynchronization [19]. An attacker may, for example,
attack is to invade into routing paths. The Multicast routing repeatedly spoof messages to an end host causing that host
protocols are targeted which reduce routing overheads. to request the retransmission of missed frames. This
The method is to quickly forward route discovery (control) degrades the ability of the end hosts to successfully send
packets by skipping processing or routing steps [18]. and receive data.
5.6 Jellyfish Attack 7 Session Layer Attacks

In a jellyfish attack the attacker first needs to intrude into In Session Layer attack an attacker attempts to hijack an
the multicast forwarding group. Then data packets established TCP session between two computers by
forwarding are delayed for some amount of time before guessing the sequence numbers and taking out one of the
forwarding them. This results in significantly high end-to- user. The counterattacks can be using encryption
end delay and thus degrades the performance of real techniques, limiting incoming traffic etc.
applications. It causes a increase in end end delay [18].
8 Application layer Attacks
5.7 Sybil Attack Application Layer attacks involve viruses, worms,
In a Sybil attack a large numbers of shadow identities are malicious codes, application abuses etc [9]. When data
generated and controlled by the malicious parties. Each being transmitted is unencrypted, it is vulnerable to
radio represents a single individual. However the sniffing as well as attacks against applications. Several
broadcast nature of radio allows a single node to pretend to additional features must be implemented within network
be many nodes simultaneously by using many different equipment to ensure against sniffing of data packets as
addresses while transmitting [18]. shown in the figure to provide additional security [21].
6 Transport Layer Attacks

Transport layer attacks may occur during authentication
and securing end-to-end communications which can be
rectified through data encryption [9].
IJSER 2011
249

ISSN 2229-5518
[11] D. Johnson and D. Maltz, Dynamic Source Routing in Ad Hoc

Wireless Networks, Mobile Computing, T. Imielinski and H. Korth, Ed.,
Kluwer, 1996.
[12] Y. Hu, A. Perrig, and D. Johnson, Ariadne: A Secure On-demand
Routing Protocol for Ad Hoc Networks, ACM MOBICOM, 2002.
[13] M. Zapata, and N. Asokan, Securing Ad Hoc Routing Protocols,
ACM WiSe, 2002.
[14] S. Ramaswamy, H. Fu, M. Sreekantaradhya, J. Dixon ,K. Nygard ,
Prevention of Cooperative Black Hole Attack in Wireless Ad Hoc
Networks , In Proceedings of the International Conference
onWireless Networks, Las Vegas, June 2003.
[15] Ebrahim Mohammed Louis Dargin,Oakland University School of
Computer Science and Engineering CSE 681 Information Security,
Routing Protocols Security in Ad Hoc Networks, Citeseer.
[16] L. Hu, D. Evans, Department of Computer Science , University of
Virginia Charlottesville, Using Directional Antennas to Prevent
Fig 4. A typical security infrastructure
Wormhole Attacks, Network and Distributed System Security Symposium
Source : [21]
(NDSS), February 2004.
[17] S. Khan, K.Loo, T. Naeem, M. Khan , Denial of Service Attacks
9 Future Work and Challenges in Broadband Wireless Networks, IJCSNS
Future work involves the study of the Identity attacks in International Journal of Computer Science and Network Security,
Wireless networks. Experimental studies will be conducted VOL.8 No.7, July 2008.
[18] V. Palanisamy, P.Annadurai, Impact of Rushing attack on
in different Wireless Technologies studying the effect of
Multicast in Mobile Ad Hoc Network , (IJCSIS) International Journal
Identity attacks in different scenarios. Possible techniques of Computer Science and Information Security, Vol. 4, No. 1 & 2, 2009.
to identify and counterattack such attacks will be suggested. [19] A. D. Wood and J. A. Stankovic, Denial of service in sensor
networks, IEEE Computer, vol. 35, no. 10, pp. 5462, 2002.
References [20] Y. Wang, G. Atterbury, B. Ramamurthy, A survey of security
[1] S. Vasudevan, K. Papagiannaki, C. Diot, J. Kurose and D. Towsley, issues in wireless sensor networks, IEEE Communications Surveys &
Facilitating Access Point selection in IEEE 802.11 wireless networks , Tutorials, vol. 8, no. 2, pp. 2-23, 2006.
Proceedings of 5th ACM Conference on Internet Measurement, 2005. [21] R. Mylavarapu, Security considerations for WiMAX- based
[2] S. Khan, K.Loo, T. Naeem, M. Hhan, Denial of Service Attacks converged network, http://rfdesign.com, August 2005.
and Challenges in Broadband Wireless Networks , IJCSNS [22] V. Abel, Survey of Attacks on Mobile Adhoc Wireless Networks,
International Journal of Computer Science and Network Security, International Journal on Computer Science and Engineering, ISSN :
VOL.8 No.7, July 2008. 0975-3397, Vol. 3 No. 2, 2 Feb 2011.
[3] V. Gunasekaran, F. Harmantzis, Emerging wireless technologies [23] V. Abel, A. Mnaouer, On the Study of the WiMAX Security
for developing countries , Technology in Society 29 2342, 2009. Threats and Current Solution Trends, Journal of the Caribbean
[4] C. Karlof and D.Wagner, Secure routing in wireless sensor Academy of Sciences, 2010.
networks: Attacks and countermeasures, Proceedings of 1st IEEE
International Workshop on Sensor Network Protocols and
Applications, 2003.
[5] J. Haartsen, BLUETOOTHThe universal radio interface for ad
hoc, wireless connectivity , Ericsson Review No. 3, 1998.
[6] A. Pirzada, C. McDonald, Establishing trust in pure ad-hoc
networks, in: CRPIT 04: Proceedings of the 27th Conference on
Australasian Computer Science, Australian Computer Society Inc.,
Darlinghurst, Australia, 2004, pp. 4754.
[7] H. Kim, Department of Electrical and Computer Engineering,
Stevens Institute of Technology, IEEE 802.16/WiMAX Security, 2006.
[8] M. Bogdanoski, P. Latkoski, A. Risteski, Borislav Popovsk, IEEE
802.16 Security Issues: A Survey, 16th Telecommunication forum,
TELFOR 2008.
[9] H Yang, H Luo, F Ye, S Lu, L Zhang , Security in mobile ad hoc
networks: challenges and solutions, IEEE Wireless Communications,
2004 Citeseer.
[10] C. Perkins and E Royer, Ad Hoc On-Demand Distance Vector
Routing,, 2nd IEEE Wksp. Mobile Comp. Sys. and Apps., 1999.
IJSER 2011
250
ISSN 2229-5518
Anomalous Hydrogen Production during

Photolysis of NaHCO3 Mixed Water
Muhammad Shahid, Noriah Bidin, Yacoob Mat Daud, Muhammad Talha, M.Inayat ullah
Abstract Production and enhancement of hydrogen on large scale is a goal towards the revolution of green and cheap energy.
Utilization of hydrogen energy has many attractive features, including energy renewability, flexibility, and zero green house gas
emissions. In this current research the production and the enhancement of hydrogen from the NaHCO3 mixed water have been
investigated under the action of diode pumped solid state laser with second harmonic of wavelength 532nm. The efficiency of the
hydrogen and oxygen yields was found to be greater than the normal Faradic efficiency. The parametric dependence of the yields
as a function of laser irradiation time, Laser focusing effect and other parameters of the electrolysis fundamentals were carefully
studied.
Index Terms Photo catalysis, Electrolysis of water, Hydrogen, Laser interaction, Electrical signals, Oxygen.
1. INTRODUCTION
W e are at the edge of an era of energy crises. The cur-

rent energy sources are not able to handle the incom-
ing huge population needs. Hydrogen is used at large
lytic process.20) But these methods are not economical and
the yields of hydrogen is not to an extent.
scale for production of ammonia, for refining the petroleum Our work on lasers has revealed the important para-
and also refining the different metals such as uranium, cop- meters which played a critical role in the enhancement of
per, zinc, tungsten and lead etc. The main source of energy hydrogen from water by laser. Most of the research work
on earth is fossil fuels which cause severe pollutions and basis on photo catalysis has carried out by flash lamps. A
cannot last for long time use. Nuclear energy is very expen- very little work is done by lasers.21) Since laser light has
sive and having disposal problems. The other sources such special properties like monochromatic, coherent, intense and
as tidal and wind schemes are not sufficient. The solar, polarize, so it was of great interest to use the laser beams as
thermal and hydral energy sources are feasible but required a an excitation source in water. The second parameter is that
lot of capital. An alternative source is water, which is cheap, the most of the work has done on light water, distilled water
clean and everlasting source of global energy and heavy water; we have used drinking water for produc-
tion of hydrogen. We have used NaHCO3 electrolyte. The
Hydrogen gas can be easily obtained by the electro- diode pumped solid state laser having a green light of wave
lysis. However, direct decomposition of water is very diffi- length 532 nm was used as an irradiation source. We inves-
cult in normal condition. The pyrolysis reaction occurs at tigated the different parameters of the laser by monitoring
high temperatures above 3700Co. 1) Anomalous hydrogen the rate of evolved gases i.e. hydrogen and oxygen. We in-
generation during plasma electrolysis was already reported. spected the dependence of hydrogen and oxygen yields as a
2-5) Access hydrogen generation by laser induced plasma laser exposure time, the effect of laser beam power and the
electrolysis was reported recently. 6-9) laser focusing effect.
Water in the liquid state has the extremely high ab- 2. Experimental Setup
sorption coefficient at a wavelength of 2.9 m.10) The effect A schematic diagram of the hydrogen reactor is shown
of generation of an electric signal, when IR-laser radiation in Figure 2.The reactor contained a glass made hydrogen
having the power density below the plasma formation thre- fuel cell having dimension 10 inch x 8 inch. Fuel cell con-
shold interacts with a water surface, was discovered by.11) tained a window for irradiation of laser, an inlet for water
The electrical signals induced by lasers were already re- and electrolyte, two outlets for hydrogen and oxygen gasses,
ported.12,13) A lot of research has been done on photo cata- an inlet for temperature probe and a D.C power supply mod-
lytic hydrogen production. The photo catalytic splitting of el ED-345B.Two electrodes steel and Aluminum were ad-
water using semiconductors has been widely studied. Many justed in the fuel cell. A CCD camera and a computer trig-
scientists produce hydrogen from water by using different gered with fuel cell for grabbing, a multimeter and gas flow
photo catalysts in water and reported hydrogen by the inte- meter are arranged with the fuel cell. The diode pumped
raction of lasers.14-18) In addition to this photolysis of wa- solid state laser with second harmonics DPSS LYDPG-1
ter has been studied using UV light.19) Solar energy has model DPG-2000 having green light of wavelength 532 nm
been used to obtained Hydrogen from water by photo cata- was placed near the fuel cell for irradiation during electroly-
IJSER 2011
251
International Journal of Scientific & Engineering Research Volume 2, Issue 4 April-2011
ISSN 2229-5518
sis of water. The DPSS laser spectrum is shown in figure

1.The drinking water 40 ml mixed with 10mL NaHCO3. In
order to start the electrolysis current was applied by D.C
source through steal and aluminum electrodes. The laser
beam from diode pumped laser was incident on water
through window of the fuel cell. The hydrogen and oxygen
produced were measured by gas flow meter. The laser beam
power was measured by a power meter model Nova
Z01500.The temperature of the water was measured by a
Temperature probe thermocouple thermometer and mercury
thermometer. The current was measured with the help of Figure 3. Oscillograms of electrical signal Peaks [7]
multimeter. The entire experimental run time was 90 mi-
nutes. The data was recorded after every minute of the run.
B
A
Figure1: DPSS laser spectrum
O2 D
H2
C PC
o
Figure 4. SEM micrographs corrosion on the electrode

DPSS Laser
Reaction Mechanism
CCD 1
H 2 O h electrolyt
e H 2 O 2 ----------- (1)
- + Camera 2
A
Power The energy deposited to the water
meter
Power
E VIt h ------------- (2)
The criteria for splitting water is
Supply DC Source
E Ed
Figure2: Schematic diagram of hydrogen reactor and
E Ed K H K O --------------------(3)
IJSER 2011
252
ISSN 2229-5518
3.2 Laser focusing effect

Where E is the total energy deposited to the water, is
the laser energy, V is the applied D.C voltage, I is the D.C It was revealed that the laser focusing effect also af-
current, t is the current rising time ,Ed is the bond dissocia- fected the yields. The experimental facts showed that when
tion energy of water, KH is the kinetic energy of hydrogen the hydrogen reactor was near the focus of the laser beam,
and KO is the kinetic energy of oxygen. the yield of hydrogen was found to be large. As long as the
distance from the focus was increased the production ob-
served to be less. Figure 6 represents this effect. The maxi-
3. Result and discussion mum yield 0.000051 cc was observed at 14 cm distance
from the focus where as minimum yield 0.000042 cc was
In order to investigate the role of electrolyte as a photo observed at 68 cm from the focal point. It was due to the
catalyst for water splitting under the influence of laser, fact that when reactor was near the focus the intensity of
various experiments were performed. It has been observed beam was large, at that point powers density became large,
that various important factors affected the yield of hydro- so yield of hydrogen found to be large. Similarly when the
gen and oxygen. distance form the focus was increased power density also
decreased, so hydrogen yield also observed to be less. The
(i) Effect of laser Power non linear behavior of the curve in the figure 6 shows the
(ii) Laser focusing effect moment of the focus point due to stirring of the water. At a
(iii) Effect of Temperature distance of 40 cm from the laser focus a sudden change in
hydrogen production was observed. This sudden change
was may be due to external disturbance produced by the
3.1 Effect of Laser power stirring effect.
The one of the important factors which affected the

product yields was Laser power. This effect is shown in 0.000052
figure 5. It has been detected that hydrogen and oxygen 0.000051
Hydrogen yield(cc)
yields increased with increase in laser input power. Initial- 0.00005

ly yields were found to be increased linearly with laser 0.000049
power. When the laser power reached at 1 watt a sudden
0.000048
increase was seen in the yields. At this point the hydrogen
yield was found to be 0.17cc where as oxygen yield was at 0.000047
0.1cc. It was due to a sharp electrical signal generated by 0.000046
the laser. This electrical signal peak had enough energy to 0.000045
overcome the bond dissociation energy of water. The hy- 0 20 40 60 80
drogen yield was found to be greater than the oxygen Distancefrom focus (cm)
yield. It was all due to the fact a lot of oxygen was used to
oxidize the electrodes. So a lot of oxygen was used in cor-
rosion process shown Figure 4.
Figure 6. Hydrogen yield versus distance from the
laser focus
0.3
3.3 Effect of Temperature
0.25
The other important factor which affected the
H2 and O2 Yields(cc)
0.2 product yields was the temperature of the water. It has

Hydrogen Yield
0.15 Oxygen yield been observed that temperature of the water rose with
time.
0.1
It was because of Joule heating effect.
0.05
H I 2 Rt (Joules)
0
Where H is heat dissipation, R is the resistance of water
0 1 2 3
Laser power(Watts)
and electrolyte and t is the current rising time. Figure 4
shows this effect. It has also observed that as long as tem-
perature raised yield of hydrogen and oxygen also in-
Figure 5. A graph of laser power versus hydrogen and creased (Figure 7). It was due to the fact that as tempera-
oxygen yields ture of water raised, bonding of the water became week
and splitting of water became prominent. The Figure 8
represents the relationship between the laser exposure time
IJSER 2011
253
ISSN 2229-5518
and the efficiency of hydrogen and oxygen yields. It was Proc. ICCF-7, Vancouver, Canada, 279, (2000).
observed that the efficiency of yields increased rapidly af- [3] Tadahiko Mizuno, Tadayoshi Ohmori, Tadashi Akimoto, Akito Taka-
ter one minute of the run and reached at 95%.After that hashi, Production of heat during Plasma Electrolysis in Liquid",J.
efficiency slightly decreased 90% and maintained this val- Appl. Phys., Vol.39, No.10 (2000).
ue throughout the run of experiment. This efficiency found [4] N.N. Il'ichev, L.A. Kulevsky, P.P. Pashinin Photovoltaic effect in water
to be greater than normal faradic efficiency. induced by a 2.92-m Cr3+: Yb3+: Ho3+: YSGG laser Quantum Elec-
tronics 35(10) 959-961 (2005).
[5] R. Mills, M. Nansteel and P. Ray Water bath calorimetric study of
excess heat generation in resonant transfer plasmas Plasma Physics,
0.45
69, 131 (2003).
0.4
0.35
[6] Muhammad Shahid,Noriah Bidin,Yacoob Mat Access Hydrogen
0.3 production by photolysis of K2CO3 mixed water. Proc. FSPGS2010,
H2 Yield (CC)
0.25 H2 yield without laser Malaysia, 179, (2010).

0.2 H2 with laser
[7] Muhammad Shahid,Noriah Bidin,Yacoob Mat ,Access hydrogen
0.15
generation by laser ablation during plasma electrolysis of drinking wa-
0.1
ter . Proc. IGCESH2010, Malaysia, (2010).
0.05
0
[8] Muhammad Shahid,Noriah Bidin,Yacoob Mat Novel hydrogen pro-
300 301 302 303 304 305 duction by laser induced electrical signals during plasma electrolysis of
Temprature(K) NaOH mixed water. Proc. of PGSRET, Islamabad, Pakistan, Pakistan,
(2010).
Figure 7. Comparison of hydrogen yields with [9] Muhammad Shahid,Noriah Bidin,Yacoob Mat ,Laser induced photo-
and without laser catalysis of electrolyte during plasma electrolysis of drinking water .
Proc. ICSM2010, Malaysia, (2010).
[10] Ilyichev N.N. et al. Kvantovaya Elektron., 35 (10), 959 (2005) Quantum
Electron, 35 (10), 959 (2005).
100 [11] S.A. Andreev, S.Yu. Kazantsev, I.G. Kononov, P.P. Pashinin, K.N. Fir-
90 sov, Temporal structure of an electric signal produced upon interac-
80 tion of radiation from a HF laserwith the bottom surface of a water col-
E ffic ienc yof yields (% )
70
umn Quantum Electronics 39 (2) 179 -184 (2009).
60
50
[12] Daniela Bertuccelli and Hctor F. Ranea-Sandoval Perturbations of
40 Conduction in Liquids by Pulsed Laser-Generated PlasmaIEEE Jour-
30 nal of Quantum electronics, Vol. 37, NO. 7, July 2001.
20 [13] Miyuki Ikeda,"Photocatalyic hydrogen production enhanced by laser
10
ablation in water-methanol mixture containg titanium(IV) oxide and
0
0 2 4 6 8 10 12
graphite silica",catalysis communications 9 (2008)1329-1333.
time(min) [14] Byeong Sub kwak,"Enhanced hydrogen production from Metha-
Figure 8. A graph of laser exposure time versus nol/water photo-splitting in TiO2 including Pd compnant" Bull Korean
efficiency of yields chem. Soc, Vol.30,No5, (2009).
[15] M.A Gondal "Production of hydrogen and oxygen by water splitting
3 CONCLUSION using laser induced photo-catalysis over Fe2O3."Applied catalysis A;
The experimental results revealed that, the diode genral 286(2004) 159-167.
pumped solid state laser with second harmonics having a [16] C. A. Sacchi, Laser-induced electric breakdown in water, J. Opt. Soc.
green light of wave length 532nm is highly efficient in Amer., vol. B 8, pp. 337345, (1991).
photo splitting of water into hydrogen and oxygen dur- [17] A. De Giacomo et al, spectroscopic "investigation of laser-water interac-
ing plasma electrolysis of NaHCO3 water. The laser pow- tion beyond the breakdown thresh hold energy"CNR.IMIP Sec Bari, via
Amendola 122/D, 70126 Bari, Italy (2007).
er, focusing effect and temperature of the water have a
[18] F. Andrew Frame,a Elizabeth C. Carroll,a Delmar S. Larsen,a Michael
significant role in enhancement of the hydrogen produc-
Sarahan,a Nigel D. Browningb and Frank E. Osterloh;First demonstra-
tion photolysis of water.
tion of CdSe as a photocatalyst for hydrogen evolution from water
under UV and visible light. DOI: 10.1039/b718796c; Berkeley, CA,
USA);(2008).
REFERENCES [19] Kestutis Juodkazis1,6, Jurga JuodkazytPhotoelectrolysis of wa-
[1] T. Mizuno, T. Ohmori and A. Akimoto Generation of Heat and Prod- ter:Solar hydrogenachievements and perspectives Optical Society of
ucts during Plasma Electrolysis,Proc. ICCF-10, (2003). America(2010).
[2] T. Ohmori and T. Mizuno, Strong Excess Energy Evolution, New [20] S.Sino,T.A.Yamamoto,R.FujimotoHydrogen evaluation from water
Elements Production, and Electromagnetic Wave and/or Neutron dispersing nano particles irridated with gamma rays/size effect and
Emission in the Light Water Electrolysis with a Tungsten Cathode, dose rate effect.Scripta matter 44(2001)1709-1712.
IJSER 2011
254
ISSN 2229-5518
[21] Kia zeng, dongka zhang "Recent progress in alkaline water electrolysis
for hydrogen production and applications"progress in energy and
combustion science (2009).
IJSER 2011
255
ISSN 2229-5518
Peripheral Interface Controller based the

Display Unit of Remote Display System
May Thwe Oo
Abstract This paper expresses about how to construct the peripheral interface controller (PIC) based the display unit of the
remote display system. The remote display system can be used to display the token number that is to know the people. It is also
intended for use in clinic, hospitals, bank, and etc. In this research, the peripheral interface controller based remote display
system is used for displaying the number and the character. The remote display system consists of two portions: display unit
and console unit. The display unit of remote display system contains the display controller, three seven-segment Light Emitting
Diode (LEDs), Diode matrix, Category display LEDs and DSUB9 connector. The display controller is controlled by the
microcontroller PIC16F873. It controls to display the token numbers. And then it can control at the diode matrix to display the
three kinds of character such as A, B, and C. The three numbers of seven-segment LEDs will display the token number from
one to 999. The diode matrix helps to display the category display LEDs. The category display LEDs will display one kind of
characters. In the research work, the category display LEDs can be displayed only three kinds of characters. For the display
unit, the DSUB9 connector applies the data that is from the console unit of the remote display system. In this research work, the
display unit works as the receiver the console unit works as the transmitter in remote display system. This paper explains about
the design, construction, testing and result of a remote display system.
Index Terms Diode Matrix, Display Unit, Light Emitting Diode, Peripheral Interface Controller, Remote Display System.
1 INTRODUCTION
T HIS display system is very useful in many applica-

tions. To advertise, to express the number, the remote
display system can be used. Because of the token
2 RELATED WORKS
number can be expressed by using the display system. 2.1 PC Basted Token Number Display
The token number may be used in banks, financial insti- An increasing number of banks are computerizing their
tutions, clinics and etc. The display unit of the remote operations. For actual field use, the display with target
display system can be controlled by the microcontroller LED arrays must be replaced. No other hardware is
PIC16F873. The token number is displayed by the large- needed, except the displays and the buffers. The absence
sized seven- segment Light Emitting Diodes (LEDs). The of any seven segment decoder chip like 7447 is to be no-
category display LED can be expressed the three charac- ticed. Because of all the decoding is done entirely by
ters. This research consists of two modules; one for the software. Time division multiplexing is another impor-
construction of hardware circuit (including interfacing tant concept. The four numbers of seven-segment dis-
circuit and LED matrix circuit) and other for how to con- plays share a common data bus. The PC places the first
trol the process by using peripheral interface controller digit on the data bus, and enables only the first seven-
(PIC). In this research, the hardware controlling program segment unit. After a delay of one millisecond, the digit is
is Assembly programming language. The circuit compo- replaced by the next digit, but this time only the second
nents consist of resistors, transistors, Integrated Circuit display unit is enabled. After displaying all the four digits
(ICs), LEDs and other accessories. The remote display in this way, the first unit is enabling again. The cycle re-
system consists of two parts. They are console unit and peats itself over and over. Another feature provided by
display unit. The display unit is used to display the num- the software is the ability to suppress leading zeros.
bers. That number is needed to know the people. That Number five is displayed as 5, rather than as 0005. This is
numbers may be the token number for using many public usually done by the hardware using the so called ripple
places. For examples, the token number may be used in blanking facility of the decoder chips. Finally, the com-
the banks and financial institutions, calling the patients puter emits a musical sound whenever there is a change
one by one at the clinics, calling the candidates at the in- in the displayed number, to attract attention. A regulated
terviewing and etc. There are many ways to display the 5-volt power supply is recommended [1].
number.
2.2 Digital Bank Token Number Display
Electronic bank token number display systems are quite
popular with bank and financial institutions nowadays.
Miss May Thwe Oo is currently pursuing doctoral degree program in These can be used by general practitioners in their clinics
school of computer science and engineering, Beihang University, China,
PH-+8613720055716. E-mail:[email protected] to call the patients one by one, and by interviewing com-
mittees to call the candidates. In most applications, there
IJSER 2011
256
ISSN 2229-5518
are two sets of displays, a small one kept inside the cabin
of the person controlling it, and a large one kept outside
for visitors. There is a keyboard for teller to enter token
numbers. The large size plasma devices and TTL chips
are being used in the constructing of bank token display
[2]. In digital bank token number display system, the dis-
play device circuit is used to control both the common
cathode LED and common anode plasma displays. A well
regulated transistorized circuit provides power to TTL
chips as well as to high voltage plasma displays.
2.3 Improvement of Display Unit

The digit display part of the display unit is made to
light up by 1 digit to suppress an electric current. The
lighting-up time of one digit becomes 1/3 and the light
falls. When little flow of the electric current occurs, the
color patches occurs to the displaying. The display data is
hold by the latch register (74LS 273). The 74LS273 which Fig. 2. Receive time chart of the receiver operation of dis-
is D-type flip-flop (D-FF) as the latch register to hold the play unit.
display of the LED is used. The eight D-FFs are housed in
3.2 Software Flowchart and Processing of the
one piece of IC. One 74LS273 is used to control one seven-
Display Unit
segment LED. It inputs the data to D terminal to make
The following flowchart shows the outline of the software
hold. First, the data for the 100th is outputted to RB6 from
processing of the display unit of the remote display sys-
RB0 and it is inputted to all the latch registers. Next, RA0
tem.
is made H and only the CK of the latch register (IC five)
for the 100th is made H level. By this, a 100th digit is dis-
played in LED1.
3 SOFTWARE IMPLEMENTATION OF DISPLAY UNIT
3.1 Receiver Operation in Display Unit

The display unit of the remote display system is based
on the PIC16F873. And it uses the asynchronous commu-
nication of PIC16F873. In asynchronous communication,
there are two operations: Transmitter operation, Receiver
operation. Fig. 1 showed the schematic diagram of receiv-
er operation of PIC16F873 in the display unit.
Fig. 3. Software flowchart for the display unit.

In the label definition, the frequently used commands
are assigned. The data area is automatically assigned
from 20h by CBLOCK directive. ENDC is used for the
ending of assignment. The label data-h is used for the
area which stores 100th data. Similarly, the label data-t is
assigned for the area which stores 10th data, data-u for
Fig.1. Schematic diagram of receiver operation of display the area which stores first data. And cdisp-p is used for
unit.
IJSER 2011
257
ISSN 2229-5518
the area which stores a category display scan position. specification. For displaying the seven-segment LEDs,
ddisp-p is to store a seven-segment LED display scan RA0, RA1, RA2 must be rewritten while d\saving this
position. rcv-p is the area which counts a received data. value.
And r-category is the workarea to use for the change of The following flowchart shows the LED control
the received category data. When making the power ON process.
of the PIC, the instruction is executed from zero address
of the program memory. When there is interruption
processing, processing is began for the address4. And
then it makes each processing jump with the GOTO in-
struction. The initialization processing is done after turn-
ing on. In this processing, all ports of the PORTA are set
output mode. And all ports of PORTB are also set to out-
put mode and are used for the segment control of the sev-
en-segment LED. RC0, RC1, RC2, RC3 are set to output
mode for the scan of the category display. RC7 pin of
PIC16F873 is the only input mode and to receive the
transferred data form the console unit.
For the USART, the most of the registers about the setting
of a USART are in bank1. Therefore, it is necessary to be
careful of the bank designation. The designation of the
asynchronous serial communication with 9600bps trans-
mission speed is done. This receiving interrupting occurs
when the data is received in the receiving buffer. When
the initialization of the USART ends, the receiving opera-
tion is immediately done. When the initialization of inter-
ruption will start, Global Interrupt Enable bit (GIE), Peri-
pheral Interrupt Enable bit (PEIE), and Timer0 Overflow
Interrupt Flag bit (T0IF) are set. To use the interruption of
transmission complete, PEIE must to set.The initialization
processing is ended. Then, it waits for the interruption
only.
As the main processing, it repeats the execution of the
same address. In the interruption process subroutine, two
kinds of interruptions are used. They are interruption by Fig. 4. Software flowchart for subroutine of LED control
the time-out of TMR0 and the interruption when the data process.
is received. The interruption of TMR0 is identified by the
TOIF bit of the INTCON register and the data receive Therefore, it makes data 0 except the category speci-
interruption is identified by the RCIF bit of the PIR1 regis- fication by the AND. By suing the AND, the result be-
ter.. The RETFIE instruction is executed at end of the in- comes 1 only when both data is 1. If the bit that fixing
terruption processing. It becomes the interruption possi- value in the instruction is 1, the contents dont change.
ble condition. When the bit that is fixing value is 0 always becomes
0. The digit specification data of the seven-segment
3.3 Software Flowchart and Processing of LED LED is set to RA0-2. Now, the calculation of the OR is
Control used. When either of the data is 1, the result becomes
The interruption occurs every two milliseconds with 1. If the contents of the bit that fixing value in the in-
TMR0. The interruption flag of TOIF should be cleared struction is 0, the contents do not change. When the bit
first. If TOIF bit is not cleared, the interruption occurs that is the fixing value is 1, the contents always be-
without waiting the desired time. Therefore, the setting of comes 1. When the digit specification is ended, the
the timer value of TMR0 is needed. As for the control of segment data will be to display. The segment data to dis-
the category display, one row is done every time it inter-
play is written in the portB. It reads the content in the
rupts in the two milliseconds. The specification of the row
received data area of each of the digits and writes in the
is done by the portC. As for the seven-segment LED con-
portB. It makes writing processing to the portB common
trol, a three digit figure is displayed by the seven-segment
and a processing step is reduced.
LED. The one digit is controlled every two milliseconds
about these LEDs. The digit specification of seven- 3.4 Software Flowchart for Data Receive Process
segments is done by RA0, RA1, and RA2. And RA3, RA4, The following flowchart is the software flowchart for sub-
RA5 of the PIC16F873 is used for the category display
IJSER 2011
258
ISSN 2229-5518
routine of data receives process. the received data is cleared. When the overrun occurs, it
stops the receiver and it must be started once again. This
process must be always done when the overrun occurs.
It clears all display data after that to find that the over-
run occurred in the process. After the overrun error is
checked, the frame error is checked. The frame is from the
start bit to the stop bit. When a stop bit is not detected
after detecting a start bit, the frame error occurs. If a nor-
mal frame is received, the error passes away. When the
frame error occurs, it clears display only. If the frame er-
ror occurs, the Framing Error (FERR) bit of the Receive
Status and Control Register (RCSTA) register is set. And
then the category data, 100th data, 10th data and 1st data
are cleared. The received position is almost cleared. The
(a) interrupt is ended.
A
After the frame error is checked, the start process is
checked. If the frame error does not occur, the start
Start?
NO process may be occurred. The category, 100th, 10th and 1st
YES Category?
NO data is sent sequentially from the console unit to the dis-
NO
Start data? YES
play unit. The information which shows the kind of the
YES
Translate to data isnt included in each data. The kind of the data is
display data
Read start data decided in order to receive. The console unit sends start
Receive position +1
SET category data
PORTA= 00CCCXXX
data first and transmits category, 100th, 10th, and 1st data
B
Receive position +1
continuously in the order. In the receiving process, it
waits for the start data first. If the start data bit is received
Clear receive it is read and the received position is set. Then, the re-
position
ceived position is incremented and the interrupt is ended.
NO
When the data except the start data comes in the start
100th ?
NO
data receiving position, the data is canceled and the in-
YES 10th ? crementation of the receiving position is not done.
Translate to
display data YES 1st ?
NO
C The category data is received behind the start data. Af-
Translate to
th
Write 100 data display data YES ter the start data is checked, the category data will be
DATA_H
Translate to checked. The category data which is sent from the console
Write 10th data display data
Receive position +1 DATA_T unit is sent in the form to use in the console unit. The bit
Receive position +1
Write 1st data
DATA_U
configuration and the bit contents of the data that is sent
from the console unit to the display unit is different. In
Clear position
the received data, bit four-six is the data which shows a
End Interrupt category. Bit six indicates A, bit five indicates B, and
bit four indicates C. And, 0 indicates lighting-up and
(b) 1 indicates going-out.
If the category data is received, it is checked to be A
Fig. 5. Software flowchart for subroutine of data receive firstly. As for the display unit, bit three-five indicates a
process: (a) for checking the overrun rrror and frame er- category. Bit three corresponds to A, bit four corres-
ror and (b) for checking the category and LED. ponds to B and bit five corresponds to C. In the dis-
play unit, 0 indicates going-out and 1 indicates light-
The data receive process makes the data receiving from ing-up. The difference is due to the difference of the
the console unit. This process is started by the data re- hardware of each unit. In the number data receive
ceive interruption by the Receive Interrupt Flag (RCIF) bit process, the 100th data, the 10th data and the 1st data are
of the PIR1 register. It is different from the other interrup- received in the order following the category data. The
tion display bit and to clear RCIF bit by the software is form of the console unit and the form of the display unit
not necessary. It is cleared by the hardware in reading the are different about these data. The bit position is the same
data which was received to RCREG. When the data is and the meaning is opposite to 0 and 1. In the console
received, the overrun error must be checked. If an over- unit, 0 is lighting-up and 1 is going-out. In the display
run condition occurs, the Continuous Receive Enable unit, 0 is going-out and 1 is lighting-up.
(CREN) bit of Receive Status and Control Register The Exclusion OR is used to reverse 0 and 1. Ac-
(RCSTA) must be cleared. Then the CREN bit is set and cording to Exclusive OR, 0 is output if being a value
IJSER 2011
259
ISSN 2229-5518
with the same data and 1 is output when the value is the PIC16F873. Fig. 6 shows the circuit diagram of the
different. The result may reverses when calculating in display unit of the remote display system.In the portion
Exclusion OR with 1. When the received data is of design of display unit of the remote display system, the
01001111, the result of execution of Exclusion OR with following circuits are including: the seven-segment LED
11111111 is 10110000. This is the value which re- control circuit, the category display control circuit, the
versed 0 and 1 of the received data. A translated val- RS232C control circuit, PIC oscillator circuit and power
ue is written in the storage area which corresponds to circuit.
IC4
+12V
each of the digits. C5 78L05 C6
R 12
When the data receiving stops in the condition that the 24

V CC
0 1
12
R1
R 13
TR2
TR1 R34 -R 46
18 G1 23
frame would be received, being normal. The receiving 19 IC2
G2 74HC154 5 6
34
4 5 R 14
R 15
TR 3 R47 -R 59
6 7 R2
processing has the possibility to make a mistake in the 12 GND
7 8
8 9
TR 4
9 10
kind of data. When the 100th data is receiving, the com- A B C D 10
23 22 21 20
11
R3
R 16
R 17
TR5 R 60 -R72
munication stopped. And then, when the start data from TR 6
R18
TR 7 R73 -R 83
the console unit is receiving, it receives the start data as R4
R 19
TR8
10th data and category data as 1st data. Because it becomes R20
R 21
TR 9 R84 -R 92
R5
start data waiting condition when 1st data is received, the TR 10
R 22
100th data, 10th data, and 1st data are canceled. So, the R6 R 23
TR11 R93 -R 100
TR 12
normal receiving is done. R 24
TR13
R 25 R 101 -R 109
R7
TR 14
R 26
4 HARDWARE IMPLEMENTATION OF DISPLAY UNIT R8

R27
TR16
TR15 R 110 -R 119
OF REMOTE DISPLAY SYSTEM R 28

R TR17 R120 -R 132
R9 29
TR18
The objective of the project is to design the display unit of R 30
TR19 R 133 -R 146
R 10R 31
the remote display system. The display unit is controlled TR20
by PIC16F873. The display unit is displayed the number R 11

R 32
R 33 TR 21 R 147 -R 156
and character that is transferred from the console unit of TR 22
the remote display system. The display unit of the remote 11 12 13 14

RC1 RC3
R 160
R157
A
RC0 RC2 5 TR23 B

display system consists of three numbers of seven- 20 V
RA3
RA4
6 R 158
TR 24 C
R 159
RA5 7
segment LEDs, diode matrix, RS232C interface,
DD
TR 25
PIC16F873, and four-to-sixteen decoder. The RS232 inter- 1

MCLR R 164 R 166
TR 26
R168
TR28 TR 30
C1 R 165 R 167 R169
R161
face is used to connect between the display unit and the 4
2 16
1
RA0 2
RA1
3 R162
TR27 TR 29 TR31
C2 C3
IC3
console unit of the remote display system. PIC16F873 con- 5
8
ADM232AAN
3
9
18
RX RA2 4
R 163
LED1 LED2 LED3
6 15
trols the three numbers of seven-segment LEDs and diode RX C4
IC1
PIC16F873
GND
matrix. a g a g a g
The remote display system consists two portions: Display 9 OSC1

RB6 27
R 170
TR 32 TR 33 TR 34 TR 35 TR36 TR 37 TR38
Unit and Console Unit. This project is to design the dis- 10

OSC2
RB5
RB4
26
25
R171
R 172
R 173
24
play unit of the remote display system. PIC16F873 is used X1
4MHz
RB3
RB2
RB1
23
22
R 174
R 175
21 R 176
to control both the display unit and the console unit of the V SS RB0
remote display system. PORTA of the PIC16F873 is used

for output mode to display the number and alphabet. All
Fig.6. Proposed circuit diagram of the display unit of the
of PORTB is to display the segments of the seven-segment
remote display system.
LEDs. A +5V is used to power of the PIC and +12V is
used to the seven-segment LEDs and diode matrix. The 4.2.1 Seven-segment LED Control Circuit
diode matrix is used to display three kinds of characters In the display unit of the remote display system, there are
such as A, B, and C.
three number of seven-segment LEDs. It is the I/O port
4.2 Design of Display Unit of the Remote Display number limitation with PIC and to make an electricity
System consumption little. Only the circuit which displays 100th
The display unit of remote display system is controlled by data is drawn in the Figure 4.2. The circuit for 10th and
the PIC16F873. There are LED selecting circuit and seg- 1st is similar too. TR27 and TR26 are the circuit which
ment selecting circuit to display the numbers with three selects a controlled LED. The drive voltage of the large-
numbers of seven-segment LEDs. And there are the row sized LED to be using this time is 12V. The drive voltage
selecting circuit and LED selecting circuit to display the for one LED is about 2V. As for one segment, five LEDs
three kinds of characters such as A, B, and C. The RS232 are connected in series. So, the drive voltage of one seg-
interface is used as the connector between the display ment becomes from 10V to 14V. The LED which was used
unit and the console unit. There is a clock circuit to drive this time is an anode common type and the anode termin-
IJSER 2011
260
ISSN 2229-5518
al which applies + voltage is common to all the segments. are used. At the LED selecting circuit, that transistor is
Because an LED selecting circuit is put in the side of the + directly controlled by the portA of PIC. If the port A of
voltage, PNP type is used for the control transistor. When PIC is 0V (0 condition), the transistor will be off and the
using NPN-type transistor for this circuit, the emitter of LED does not light-up. When the portA of PIC is 5V (1
the transistor is connected with the anode terminal of the condition), the base electric current flows through the
LED. In the case, the electric current control for the base transistor. And the electric current through the LED and
becomes difficult. the LED will be light-up.
At the segment selecting circuit, the NPN transistor be-
comes off when the portB of PIC is making 0V (0 condi-
tion). So, the base electric current of PNP transistor does
TR26
not flow and the PNP transistor will off. The electric cur-
rent does not through the segment of seven-segment LED
and the segment does not light-up. If the portB of the PIC
is 5V (1 condition), the NPN transistor will be on. The
TR27
electric current flows through the base of the PNP transis-
tor and that transistor will be on condition. And the elec-
tric current flows through the segments and the segment
TR32 lights up. The PNP transistor is not directly by the PIC.
4.2.2 Category Display Control Circuit
Fig. 7. Schematic diagram of 7-segment LED control cir- In the display unit, the category is displayed with the di-
cuit. ode matrix. Fig. 8 shows the category display control cir-
cuit. That consists of the row selecting circuit and the LED
The TR27 is used for the output of PIC. This is because
the voltage which is applied to the I/O port of PIC is li- selecting circuit. A circuit like seven segments is used
mited to +5V. When making TR26 ON, it makes the base about the drive circuit for the category display. A catego-
voltage of TR26 less than +12V (About11V). The electric ry character is one character but is displayed by the LED
current flows through the base with this. When making matrix. An LED matrix is composed of 11 lines to the di-
TR26 OFF, it makes base voltage +12V. So, about +12V rection of the side (Row) and is composed of 13 lines to
the longitudinal (Column). Alighting-up control is done
voltage is applied to the base of TR26. The base of TR26
can not be directly driven by PIC. When making RA0 of every row. The row selecting circuit controls to pass an
PIC 0V (0 conditions), TR27 becomes OFF. Therefore, the electric current to the LED of the row which was specified
base electric current of TR26 does not flow and TR26 be- by PIC. Because there are 11 rows, the direct control by
comes OFF too. That is, LED1 does not light up. When PIC is difficult. It is to limit the number of the I/O ports.
A control signal is developed from the four-bit signal of
making RA0 of PIC +5V (1 condition), the base electric
current flows through TR27 and TR27 becomes ON. PIC to 11 signals by the decoder IC. The operation of the
When TR27 becomes ON, the electric current flows row selecting circuit is similar to the case of seven seg-
through the base of TR26 and TR26 becomes ON condi- ments.
tion. By this, +12V are applied on the anode of the LED
and the LED becomes the condition about which it is
possible to light up. TR1
The segment selecting circuit drives a lit segment. A seg-
ment selecting circuit is put between the LED and the
grounding. So, the transistor for the control can be direct- TR2
ly driven by PIC. In the left figure, a circuit for the "a"
segment is drawn. The other segment control circuit is
similar too. When making RB6 of PIC 0V (0 conditions), TR23
TR32 becomes OFF. Therefore, the electric current does TR24
not flow through the "a" segment of the LED and the "a" TR25
segment does not light up. The base electric current flows
through the base of TR32 when RB6 of PIC is +5V (1 con-
dition). With this, the electric current flows through the Fig. 8. Schematic diagram of the category display control
"a" segment and the "a" segment lights up. In this re- Circuit.
search, the common cathode seven-segment LED is used. The LED selecting circuit is the circuit which drives a lit
So, the LED selecting circuit used only one NPN transis- LED in the selected row. The character to display with the
tor for each seven-segment LED. For the segment select- circuit this time is three kinds. The kind of the character is
ing circuit, one PNP transistor and one NPN transistor controlled by RA3, RA4 and RA5 of PIC. The LED which
IJSER 2011
261
ISSN 2229-5518
is lit up every character is different. When seeing from the

LED, there are one which lights up by all character, one
which lights up only in case of specific character and so
on. It is the diode of the LED selecting circuit that is con-
trolling this. In case of the circuit of the figure on the left,
the case of the LED which lights up by all character is
shown. When making an LED light up only when TR1 is
ON, it makes a diode only the one which is connected
with the collector of TR23. It decides a displaying charac-
Fig. 10. Schematic Diagram of the PIC Clock Circuit
ter pattern in the LED matrix and it decides the combina-
tion of these diodes every LED. 4.2.5 Power Circuit
4.2.3. RS232C Control Circuit The power to operate the display unit is supplied from
the console unit through the cable. The DC +12V is made
The RS232C interface is used for the information
with AC +220V by the switching power circuit. The sup-
transfer between the console unit and the display unit.
plied voltage from the console unit is +12V. The +12V can
This interface is used for an interface with the modem.
be used for the LED displaying. The power of PIC needs
About 9V is used for signal voltage on the connection
the +5V. A three-terminal regulator is used for the voltage
cable with the console unit and the long distance trans-
change from +12V to +5V. The +5V is used for PIC16F873
mission is possible comparatively. In the standard, the
and decoder IC. The regulator can be used as the 100mA
cable with about 15m length can be used. Actually, it is
type regulator.
possible to use for more distance. It depends on the con-
dition of the cable. The +5V signal of PIC16F873 can be
directly transmitted without using the RS232C interface if
being a short distance. The purpose that this IC is used
this time is to attempt to use the RS232C device and to
secure a distance. The ADM232AAN has the DC-DC con-
verter which makes the voltage of 10V with +5V power.
The external circuit is very simple.
Fig. 11. Schematic Diagram of the Power Circuit
4.3 Test and Result of the Display Unit of the

Remote Display System
The peripheral interface controller based the remote dis-
play system is shown in Fig. 12. The step-down transfor-
mer is used to change from the AC 220V to the 12V. The
Fig. 12 is the figure that is before the power supply is off
and the data is not being sent from the console unit to the
display unit of the remote display system.
Fig. 9. Schematic diagram of the RS232C control circuit.
4.2.4 PIC Oscillator Circuit

To make do the operation of PIC, a clock generator is
needed. An oscillation had within PIC, so, it puts a vibra-
tor outside only. As the vibrator, the crystal oscillator can
be used because of the precision oscillation. However, at
the circuit this time, a ceramic vibrator (Resonator) is
used because it does not need the precision oscillation.
The maximum clock frequency of PIC16F873 is upto
20MHz. However, at the equipment this time, because it
does not need high-speed operation, the oscillation fre-
quency is 4MHz. A transmission speed is calculated by
4MHz. So, when using this equipment at the clock fre-
quency which is not 4MHz, reconsideration is needed.
Fig. 12. The PIC based remote display system.
IJSER 2011
262
ISSN 2229-5518
When the number 450 and the character A is sent from

the console unit, the data that is sent from the console
unit is shown in Fig. 13.
Fig. 15. Testing the received voltage level at the display

unit.
4 CONCLUSION
This research describes about the display unit of the re-
mote display system. The peripheral interface controller
Fig. 13. Testing the receiving data both console unit and based remote display system can be used to inform the
display unit in remote display system. people. In this project, the display unit is constructed, one
of the nearest, most versatile and most useful LED array,
to view alphabetic character and numeric characters. In
The Fig. 14 shows the display portion of the remote this research, the three kinds of characters and the three
display system. digits numbers are displayed. The PIC16F873 controls the
numbers and the characters. The display unit of peripher-
al interface controller based remote display system is de-
signed to construct using with the devices which can buy
easily in the market, easy to use and familiar with a per-
son who interested in electronics field. This system is de-
signed for English language capital letters and numbers.
This thesis is not only applied for the educational aid for
the beginner to learn the design, programming and de-
velopment of the applications which use the PIC but also
applied for displaying the number in many fields such as
clinic, bank and financial institutions, interviewing com-
mittees to call the candidates and etc.
As further extension, the remote display system is
used 15m length between the console unit and the display
unit in this research. For more distance, the various kinds
of RS232 interface can be used. In large area of display, it
can be used the largest size of seven-segment LED. By
this way, the resolution of the display is smooth. Instead
of LED, LCD can be used in this research. In this research,
the digit display part of the display unit is made to light
up by one digit to suppress an electric current. It can be
Fig. 14. The display unit with the data receiving process.
made to light up the digits at the same time. It can be
used to hold the display data by latch register (74LS273).
The supply voltage of the display unit is supplied from
The latch register is placed between the peripheral inter-
the console unit of the remote display system. Fig. 15
face controller and the segment selecting circuit. The re-
shows the testing of received voltage level at the display
mote display system can be widely used in many fields
unit.
that need to give information to users or partners. To de-
velop the display technique and devices, people should
IJSER 2011
263
ISSN 2229-5518
study this technology. The development of this project

intended the display unit of the remote display system to
be used in various fields.
ACKNOWLEDGMENT
Firstly the author would like to thank her parents for their best
wishes to join the Ph.D research. The author would like to ex-
press the heart-felt gratitude to Daw Atar Mon for her leader-
ship and advice, Daw Khin Sandar Tun for guidance in her re-
search and U Tun Tun Win for suggestions on how to design
the remote token display system. The author greatly expresses
her thanks to all persons who will concern to support in prepar-
ing this paper and her research.
REFERENCES
[1] V.Rajarama and Mohan Ingle, Electronic Project Volume-16, An EFY
Enterprise Publication, 1995.
[2] BEL Application Lab, Electronic Project Volume-8, An EFY Enter-
prise Publication, 1987.
[3] Inoue, S, Remote Display System, PIC Circuit Gallery, October 2005.
[4] Data Sheet PIC16F873 28/40-Pin 8-bit CMOS FLASH Microcontrol-
lers, <http://www.microchip.com>
[5] +5V-Powered, Multichannel RS-232 Drivers/Receivers.
<http://www.maxim.ic.com/packages>
[6] Data Sheet of 74HC/HC154 4-to-16 Line Decoder/Demultiplexer,
September, 2004. <http://www.philips.com>
[7] http://www.Datasheet 4u.com
[8] Thomas L. Floyd, Electronic Devices I, 4th Edition, Prentice Hall
International, Inc, 1996.
[9] Thomas L. Floydm, Electronic Devices II. 4th Edition, Prentice Hall
International, Inc, 1996.
[10] Thomas E.Kissell, Industrial Electronics, Printed in the Republic of
Singapore, Prentrice-Hall International Editions.
[11] Thomas C.Hayes Paul Horowitz, Student Manual for the Art of Elec-
tronic, Printed in Great Britain at the University Press, Cambridge.
[12] Nashelsky, L.and Boylestad, R.L., Electronic Devices and Circuit
Theory. 8th Edition, U.S.A. Prentice Hall, Inc. 2002.
[13] Boylestand, R., Electronic Devices and Circuit Theory, Fifth Edition,
Printed by Prentice-Hall International, Inc.
[14] http://www.interq.jp.au.com
[15] http://www.piclist.com
[16] http://www.st.com
[17] http://www.datasheetarchive.com
IJSER 2011
264
ISSN 2229-5518
Training and Analysis of a Neural Network

Model Algorithm
Prof Gouri Patil
AbstractAn algorithm is a set of instruction pattern given in an analytical process of any program/function-ale to achieve
desired results. It is a model-programmed action leading to a desired reaction. A neural network is a self-learning mining model
algorithm, which aligns/ learns relative to the logic applied in initiation of primary codes of network. Neural network models are
the most suitable models in any management system be it business forecast or weather forecast. The paper emphasizes not
only on designing, functioning of neural network models but also on the prediction errors in the network associated at every step
in the design and function-ale process.
Index Terms Input , Neural Network, Training, Weights
1 NEURAL NETWORK A MINING MODEL
A Neural Network algorithm could contain multiple

networks, depending on the number of columns
used for both input and prediction, or that are used
transfer function wide enough to accept input in any nu-
meric range and specific enough to give output in the
desired range. The range and limit of inputs should be
only for prediction. The number of networks a single min- predefined. Saturation/threshold limit for inputs could
ing model contains depends on the number of parameters then be pre-defined, so that inputs in a way are pro-
connected by the input columns and predictable columns grammed to give the desired outputs or those inputs or
the mining model uses. Neural network function-ale is a functions incompetent of giving the desired outputs are
mimic of human neural interconnections and memory. straight away ruled out in a programmed function-ale
Human brain or memory comprises of an average of network model. Networks are built in a manner that they
about ten billion neurons functioning in a network syn- induce a stepwise logical action called the learning rule.
chronization -- and every single neuron is, on average Numeric functional values for the inputs are thus prede-
connected to several other neurons ( may be around a fined and scaled but for some non-numeric nominal val-
thousand) indirectly or directly to the central neural mass, ues as male or female, yes or no the network functions are
the brain. By way of these connections and interconnec- divided as different links in the network and graded on
tions, nerves send and receive messages as packets of an another chain of Numeric functional values as inputs
energy called as impulses. One very important feature of like (0 and 1) and thus branching the network in two si-
human neurons is that they don't react immediately to the mulating paths working on the same predefined thre-
reception of energy called impulse. Instead, they sum shold value to decide the final output of the process. The
their received energies, and then, they send their own input data for training is the fundamental factor of a
quantities of energy to the other associated neurons only neural network, as it gives required information to "dis-
when this sum reaches a certain critical threshold the cover" the optimal threshold operating point and the re-
neurons trigger and respond back as signals or packets of sult or output of the network. If the output of the network
energy called impulse. Brain learns by adjusting the is known and weighed inputs are adjusted to reach the
number and strength of these connections and gives a desired output then the network training is called super-
desired response or output in terms of polarization or revised neural network. When the output is not defined or
polarization of neurons and difference in the energy le- known, as in the case of sale or stock prediction the net-
vels. Neural networks also work on the same principle as work is an unsupervised network, which is generally
our brains work, they respond to a threshold level of programmed to trigger for all hit value of the input. In
input signal often calculated as weights in the network case of unsupervised networks for hidden neurons, input
system. values of the network may adjust and interpret differen-
tially in different situations and the network ripens or
learns based on the adjustments made in the input values
2 INPUT VALUES IN A NEURAL NETWORK MODEL
of hidden neural layer.
As all inputs and outputs of neural networks are numeric,
primary task in designing a neural network is to define a

Prof Gouri. A.Patil is currently working as a faculty for Bioinformatics with
Welingkar Institute of Management, Development and Research, Matunga
Mumbai -Central, India, contact-+91 9833025213
IJSER 2011
265
ISSN 2229-5518
Wo=bk
Input (bias) additional benefit of having an extremely simple
X0 = +1 derivative function for back propagating errors through
the neural network.
X W
0 0 Out
X
1
W
1
Ac- Net
Sum- tiva- Out
ming tion put
W up Func
X 2
juncn -tion
2
Out = 1/1+e-Net
Sigmoid function
W
X n
n
Out
Wei-
Input ghts
sig-
nals
Thre
shol
d
Net
3 FUNCTION-ALE OF A NEURAL NETWORK MODEL

The adjustments made in the input values of the hidden Out = tanh (Net/2)
neural layer of an unsurpervised network is nothing but tanh function
calculations in neural network in terms of adjustments of
their weights at the summing up junction and the
decisions made at the activation point by defining an
activation function value specific for every condition and Typically the weights in a neural network are initially
value. Decisions of the factors include calculations with set to small random values; this represents that the net-
loops and functions. Initially some time lapse as a work knows nothing to begin with. As the training
process proceeds, these weights would converge to val-
function is used to process the data and "winner input in
ues allowing them to perform a useful computation. Thus
the time lapse of the model takes all the credit in the first
it can be said that the neural network begins with know-
step, where the neuron with the highest value from the ing nothing and moves on to gain admirable real world
calculation fires and takes a value 1, and all other neurons application.
take the value 0.
Activation function is an important function, which de-
It is advisable that the sigmoid curve is used as a transfer
cides the maturation and output of a neural network. Ac-
function because it has the effect of "squashing" the inputs tivation functions of the hidden values in the neural net-
into the range [0,1]. The sigmoid function has the work could introduce non linearity and desired matura-
IJSER 2011
266
ISSN 2229-5518
tion of a neural network otherwise the network would insulin in correct amount and at required times so that
have been just a plain mathematical algorithm without body learns to metabolize efficiently with inefficient or
logical application. For feedback or feed forward learning week pancreas. Ageing is one of the prime factor for all
of network, the activation function should be differentia- metabolic disorders and so it goes for diabetes, our pan-
ble as it helps in most of the learning curves for training a creas ages right along with us, and doesn't pump ade-
neural network like the bounded sigmoid function , the quate levels of insulin as efficiently as it did when we
logistic tanh function with positive and negative values were younger. Also, as our cells age, they become more
or the Gaussian function. Almost all of these nonlinear resistant to insulin (the carrier of glucose for the cells) as
activation function conditions assure better numerical well. Modern sedentary lifestyle is damaging our healths
conditioning and induced learning. and is a prime responsive factor for growing obesity
Networks with threshold limits with out activation func- problems." being obese or overweight is one of the prime
tions are difficult to train as they are programmed for factors for increasing level of glucose in the blood causing
step wise constant rise or fall of weight. Where as a sig- diabetes. Obesity increases fat cells in the body and fat
moid activation functions with threshold limits makes a cells lack insulin or glucose receptors compared to muscle
small change in the input weight produces change in the cells, thus increasing fat in the cells is an indirect call to
output and also makes it possible to predict whether the diabetes, exercising and reducing fat in the cells can act as
change in the input weight is good or bad. an alert to stay away from this disorder. Eating less of fat
and enough fibre and complex carbohydrates compared
In the activation function training, Numerical condition is
to simple carbohydrates could contribute to reduce the
one of the most fundamental and important concepts of
risk of diabetes. A survey research on 20 individuals for
the algorithm, it is very important that the activation
risk to diabetes was conducted using only 4 non-clinical
function of a network algorithm is a predefined numeric
parameters. A supervised network was built and trained
condition. Numerical condition affects the speed and
to give an output equivalent to associated risk to diabetis
accuracy of most numerical algorithms. Numerical condi-
with a predefined threshold limit to reach safe non di-
tion is especially important in the study of neural net-
abetic level with maximum iterations possible at the acti-
works because ill-conditioning is a common cause of slow
vation function to reach the out put, which of course was
and inaccurate results from many network algorithms.
not fixed but was predefined to reach to zero risk to di-
Numerical condition is mostly decided based on condi- abetes.
tion number of the input value, which for a neural net-
Sample data collected
work is the ratio of the largest and smallest eigenvalues.
of the Hessian matrix. The eigenvalues of inputs are the No Age Stress level/ BMI Obesity
squares of the singular values of the primary input and Kind of work
Hessain matrix is the matrix of second-order partial de- 1 30 Stressful labour 30 Over Weight
rivatives of the error function with respect to the weights 2 24 Sedentary work 29 Over weight
and biases. 3 23 Sedentary work 32 Obese
4 35 Sedentary work 40 Highly obese
5 40 Minimum work 33 Obese
4 MODEL NEURAL NETWORK FOR A SAMPLE SIZE 6 45 Minimum work 34 Obese
OF 20 INDIVIDUALS TO TEST RISK TO DIABETIS. 7 43 Minimum work 22 fit
8 50 Minimum work 26 Slightly over weight
Type 2 diabetes, is a non-neonatal kind of diabetes that
9 55 Maximum work 20 fit
develops in the later stages of life due varied reasons, is
far more severe and chronic than type 1 or neonatal di-
abetis in its destructive metabolic effects on the body. An
12 38 Minimum work 30 Obese
associated gland called pancreas in our body synthesis
hormone called insulin required to metabolize, break-
14 29 Stressful labour 17 fit
down and capture glucose for every cell of the body to
15 30 Stressful labour 18 fit
synthesize energy in the form of energy rich molecule
16 34 Sedentary work 20 fit
called ATP (Adenosine tri- phosphate). Metabolism of
glucose is an indispensable process for the body as this
metabolism generates energy for the cell at micro and
body at the macro level. There are varied factors response
20 43 Maximum work 18 Fit
for the onset of this metabolic disorder called diabetes,
control of some of these factors could induce synthesis of
IJSER 2011
267
ISSN 2229-5518
5 12 nn = n1+ n2+ 4 Acute risk to

No We- Wei- We- Weig-ht Summ- diabetes
No Input Parameter Range Predefined n3+ n4- n1-3
ight ght of ight of input ing up
Weight 6 12 nn = n1+ n2+ 4 Acute risk to
Input input of obesity Junction
1 Age (0-20) 1 n3+ n4- n1-3 diabetes
age stress input
2 Age ( 20-30) 2 7 9 nn = n1+ n2+ 1 No risk
level BMI
3 Age ( 30-40) 3 n3+ n4- n1
4 Age (n340-60) n4 8 11 3 Acute risk to
n 4
nn = n1+ n2+
n1 n2 n
51 Stressful labor Stressful labor 0 n3+ n4- n1-2 diabetes
3 0 2 3 8
62 Maximum 9 8 1 No risk
3 work Maximum work 1 nn = n1+ n2+
2 2 3 9
73 Minimum n3+ n4- n1+1
2 3 work 3Minimum 4 work 10 2
84 Sedentary 10 8 nn = n1+ n2+ 1 No risk
3 3 work 4Sedentary 5 work 15 3
95 No work No work n3+ n4- n1+1
4 2 3 3 12 4
10 BMI (3 1 -25) 3 11 6 nn = n1+ n2+ 1 No risk
6 4 2 12 1 n3+ n4- n1+2
11
7 BMI
4 3 (1 25-30) 1 9 2
12 BMI (30-35) 12 12 nn = n1+ n2+ 4 Diabetic
8 4 3 2 2 11 3 n3+ n4- n1-4
13
9 BMI
4 2 1(35-40) 1 8 4
14 Obesity test Fit 1 13 5 nn = n1+ n2+ 0 No risk
10 4 2 1 1 8 n3+ n4
15
11 Obesity
3 test
1 1Slightly over
1 weight
6 2
16 Obesity test Over Weight 14 4 nn = n1+ n2+ -1 0 No risk
12 3 2 3 4 12 3 n3+ n4+1
17
13 Obesity
2 test
1 1Obese 1 5 4
15 5 nn = n1+ n2+ 0 No risk
18
14 Obesity
2 test
0 1Highly obese
1 4 5
n3+ n4
15 3 0 1 1 5 16 8 nn = n1+ n2+ 1 No risk
16 3 3 1 1 8 n3+ n4- n1
17 3 3 1 1 8 17 8 nn = n1+ n2+ 1 No risk
18 3 3 1 1 8 n3+ n4- n1
19 4 2 1 1 8 18 8 nn = n1+ n2+ 1 No risk
20 4 1 1 1 7 n3+ n4- n1
Predefined Weights for data 19 8 nn = n1+ n2+ 1 No risk
Out put = Threshold limit and risk to diabetes n3+ n4- n1+1
Condi- Maximum Risk to Diabetis 20 7 nn = n1+ n2+ 1 No risk
tion No Threshold limit n3+ n4- n1+2
1 5 No risk to Diabetis
2 6 Minimum risk to
Diabetes This is a very simple prototype of a network, where every
3 7 or more Acute risk to diabetes layer of inputs leads to an output. There is no secondary
Network training and summing up Junction training and learning or firing rule developed in this net-
work, however every network begin with as simple learn-
Actiavtion function and Iterations to reach the final ing single neuron like this and mature to most neural and
output from Summing up Junction complex forms. As neural links in the network increases
No Summ- Activation Itera- Output- complexity in training and analyzing the network also
ing up function tions/tr Predict increases, at the latest with improper links, training and
Junction n n = n1+ n2+ -aining how/when; analysis most networks often fails. The next section dis-
n3+ n4- n 1/n2/ n3 discard nn 5 cusses the common causes for failure of maturing net-
/n 4- 1/2/3 such works and ways to combat these issues with defining
n n as n n 5 open/flexible activation function along with threshold
1 8 nn = n1+ n2+ 1 regular exer- limits to fire an output.
n3+ n4 - n1 cise required
2 9 nn = n1+ n2+ 2 Acute risk to
n3+ n4 diabetes
- n1- 3
5 SOME COMMON CAUSES AND REMEDIES FOR ILL
3 10 nn = n1+ n2+ 4 Diabetic CONDITIONING IN A NETWORK
n3+ n4- n1- 5
1. Low coefficient of variation (standard deviation
4 15 nn = n1+ n2+ 4 Diabetic
divided by the mean) of input variables causes
n3+ n4- n1- 7
IJSER 2011
268
ISSN 2229-5518
great difference in the numeric functions. Such tific Digital Literature Library
an issue probably could be minimized by sub- [7] Genetic Programming in Data Mining for Drug Dis-
tracting the mean of each input variable from the covery - Langdon, Barrett (2004) CiteSeer.IST Seintif-
variable. ic Digital Literature Library
[8] Making Indefinite Kernel Learning Practical - Miers-
2. If the variances among input variables increase wa (2006) (Correct)
measurably, problem can be cured by dividing [9] Evolutionary Learning with Kernels: A Generic Solu-
each input variable by its standard deviation. tion for Large.. - Mierswa (2006) CiteSeer.IST Seintif-
ic Digital Literature Library
3. High correlations among input variables. This [10] Combining Decision Trees and Neural Networks for
problem could be cured by ortho-normalizing Drug.. - Langdon, Barrett, etal. (2002) CiteSeer.IST
the input variables using Gram-Schmidt, SVD, Seintific Digital Literature Library
principal components. The orthogonal compo- [11] Diversity in Neural Network Ensembles - Gavin
nents in this case should be standardized before Brown To (2003)
taking as input value. Single orthogonal compo- [12] Linear Equality Constraints and Homomorphous
nents which fail standardization should be re- Mappings in PSO - Christopher Monson CiteSeer.IST
moved from the network as they cannot trained Seintific Digital Literature Library
or accepted in the network. Training them in the
long process mail cause network failure.
4. If some hidden training units in the network be-

comes unavoidable and in the long run, the net-
work may become saturated with many hidden
units. This issue could be handled by using
weight decay or similar regularization methods
on the input-to-hidden weights, at the risk of less
accuracy in learning discontinuities or steep
areas of the target function.
5. Low coefficient of variation for activation values

of a hidden unit.-this problem can often be ame-
liorated by using a hidden unit activation func-
tion also with an output range of (-1,1), such as
tanh, instead of (0,1) as in the logistic function.
6 REFERENCES
[1] Neural Network design, Martin T Hagan, Howard B
Demuth, Mark Beale, 1996 Cengage Learning India
Pvt Ltd, ISBN-13 978-81-315-0395-9W.-K. Chen, Li-
near Networks and Systems.
[2] Ill-Conditioning in Neural Networks, Warren S.
Sarle, SAS Institute Inc., Cary, NC, USA Sep 5,
1999Copyright 1999 by Warren S. Sarle, Cary, NC,
USA
[3] The Machine Learning Dictionary for COMP9414
[4] Analytics Consulting, Prasanna Parthasarathy, Man-
agement Journal Bharathidasan Institute of Manage-
ment, Trichy , 324-327
[5] Comparison of Artificial Neural Network and Logis-
tic Regression Models for Prediction of Mortality in
Head Trauma Based on Initial Clinical Data, Bio-Med
Central Feb 2005
[6] R. Roy.. - Furuhashi And Homann, Soft Computing
and Industry Recent Applications, ,CiteSeer.IST Sein-
IJSER 2011
269
International Journal of Scientific & Engineering Research Volume 2, Issue 4, April -2011
ISSN 2229-5518
Wireless Sensor Network: A Review on Data

Aggregation
Kiran Maraiya, Kamal Kant, Nitin Gupta
Abstract Data aggregation is very crucial techniques in wireless sensor network. Because with the help of data aggregation we
reduce the energy consumption by eliminating redundancy. When wireless sensor network deployed in remote areas or hostile
environm ent. In the wireless sensor network have the most challenging task is a life time so with help of data aggregation we can
enhance the lifetime of the network .In this paper we discuss the data aggregation approaches based on the routing protocols, the
algorithm in the wireless sensor network. And also discuss the advantages and disadvantages or various performance measures of
the data aggregation in the network.
Index Terms Wireless sensor network, data aggregation, architecture, Network Lifetime, Routing, Tree, Cluster, Base Station.
1 INTRODUCTION
T HE wireless sensor network is ad-hoc network. It

consists small light weighted wireless nodes called
sensor nodes, deployed in physical or environmental
condition. And it measured physical parameters such as BS
sound, pressure, temperature, and humidity. These sensor
nodes deployed in large or thousand numbers and
Target Sensor node Sensor field Internet
collaborate to form an ad hoc network capable of reporting
to data collection sink (base station). Wireless sensor
network have various applications like habitat monitoring,
building monitoring, health monitoring, military survival
lance and target tracking. However wireless sensor User
network is a resource constraint if we talk about energy,
computation, memory and limited communication Figure 1 Architecture of the Sensor network
capabilities. All sensor nodes in the wireless sensor A sensor nodes that generates data, based on its sensing
network are interact with each other or by intermediate mechanisms observation and transmit sensed data packet
sensor nodes. to the base station
(sink). This process basically direct transmission since the
base station may located very far away from sensor nodes
needs.
More energy to transmit data over long distances so that a
better technique is to have fewer nodes send data to the base
station. These nodes called aggregator nodes and processes
called data aggregation in wireless sensor network.
2. CLUSTERING IN WSN
Sensor node are densely deployed in wireless sensor
network that means physical environment would produce
very similar data in close by sensor node and transmitting
such type of data is more or less redundant. So all these
facts encourage using some kind of grouping of sensor
Kiran Maraiya is currently pursuing masters degree program in Computer nodes such that group of sensor node can be combined or
Science and engineering in NIT Hamirpur, India, PH-+91 9318583266. E- compress data together and transmit only compact data.
mail: [email protected] This can reduce localized traffic in individual group and
Kamal Kant has joined as Lecturer in ASET,Amity University, Noida
(U.P.) , India, PH-+919718281158 E-mail: [email protected] also reduce global data. This grouping process of sensor
Nitin Gupta has joined as Assistant Professor in NIT Hamirpur, India, nodes in a densely deployed large scale sensor node is
PH-+911972254416 E-mail: [email protected]
IJSER 2011
270
ISSN 2229-5518
known as clustering. The way of combing data and 2. Heterogeneity in a network, it means user can put some
compress data belonging to a single cluster called data power full nodes, in term of energy in the network which
fusion (aggregation). can behave like cluster head and simple node in a cluster
Issues of clustering in wireless sensor network:-
work as a cluster member only.
1. How many sensor nodes should be taken in a single
cluster. Selection procedure of cluster head in an individual Many protocols and algorithm have been proposed which
cluster. deal with each individual issue.
3. DATA AGGREGATION
In typical wireless sensor networks, sensor nodes are burden before they are transmitted to the base station or
usually resource-constrained and battery-limited. In order sink. The wireless sensor network has consisted three types
to save resources and energy, data must be aggregated to of nodes. Simple regular sensor nodes, aggregator node
avoid overwhelming amounts of traffic in the network. and querier. Regular sensor nodes sense data packet from
There has been extensive work on data aggregation the environment and send to the aggregator nodes
schemes in sensor networks, The aim of data aggregation is basically these aggregator nodes collect data from multiple
that eliminates redundant data transmission and enhances sensor nodes of the network, aggregates the data packet
the lifetime of energy in wireless sensor network. Data using a some aggregation function like sum, average, count,
aggregation is the process of one or several sensors then max min and then sends aggregates result to upper
collects the detection result from other sensor. The collected aggregator node or the querier node who generate the
data must be processed by sensor to reduce transmission query.
packet transmission. And also save energy of the sensor

BS BS
node in the wireless sensor network. With the help of data
aggregation we enhance the lifetime of wireless sensor
network. Sink have a data packet with energy efficient
manner with minimum data latency. So data latency is very
important in many applications of wireless sensor network
7
such as environment monitoring, health, monitoring,
8 where the freshness of data is also an important factor. It is
critical to develop energy-efficient data-aggregation
algorithms so that network lifetime is enhanced. There are
1 2 3 several factors which determine the energy efficiency of a
4 6
5 sensor network, such as network architecture, the data-
aggregation mechanism, and the underlying routing
Figure 2 Data aggregation model and Non data aggregation model protocol. Wireless sensor network has distributed
processing of sensor node data. Data aggregation is the
It can be the base station or sometimes an external user technique. It describes the processing method that is
who has permission to interact with the network. Data applied on the data received by a sensor node as well as
transmission between sensor nodes, aggregators and the data is to be routed in the entire network. In which reduce
querier consumes lot of energy in wireless sensor network. energy consumption of the sensor nodes and also reduce
Figure 2 contain two models one is data aggregation model the number of transmissions or length of the data packet.
and second is non data aggregation model in which sensor Elena Fosolo et al in [5] describe In network aggregation is
nodes 1, 2, 3,4,5,6 are regular nodes that collecting data the exclusive process of collecting and routing information
packet and reporting them back to the upper nodes where through a multi hop network. Processing of data packet
sensor nodes 7,8 are aggregators that perform sensing and with the help of intermediate sensor nodes. The objective of
aggregating at the same time. In this aggregation model 4 this approach is increasing the life time of the network and
data packet travelled within the network and only one data also reduces resource consumption. There are two type of
packet is transmitted to the base station (sink). And other approach for in network aggregation. With size reduction
non data aggregation model also 4 data packet travelled and without size reduction .In network aggregation with
within the network and all data packets are sent to the size reduction. It is the process in which combine and
base station(sink), means we can say that with the help of compressing the data received by a sensor node from its
data aggregation process we decrease the number of data neighbors in order to reduce the length of data packet to be
IJSER 2011
271
ISSN 2229-5518
sent towards the base station. Example, in some Latency: Latency is evaluate data of time delay experiences
circumstance a node receives two data packets which have by system, means data send by sensor nodes and received
a correlated data. In this condition it is useless to send both by base station(sink).basically delay involved in data
data packets. Then we apply a function like MAX, AVG, transmission, routing and data aggregation.
and MIN and again send single data packet to base station. Communication overhead: It evaluates the communication
With help of this approach we reduce the number of bit complexity of the network fusion algorithm.
transmitted in the network and also save a lot of energy. In Data accuracy: It is a evaluate of ratio of total number of
network aggregation without size reduction is defined in reading received at the base station (sink) to the total
the process of data packets received by different neighbors number of generated. There are different types data-
in to a single data packet but without processing the value of aggregation protocols like network architecture based data-
data. This process also reduces energy consumption or aggregation protocols, network-flow-based data-
increase life time of the network. aggregation protocols and quality of service (QOS)-aware
data-aggregation protocols designed to guarantee QOS
3.1 Advantage and Disadvantage of Data metrics. Here network architecture based protocols are
aggregation in wireless sensor network described in detail.
Advantage: With the help of data aggregation process we
3.3 Impact of data aggregation in wireless sensor
can enhance the robustness and accuracy of information
network
which is obtained by entire network, certain redundancy
exists in the data collected from sensor nodes thus data In this paper we discuss the two main factors that affect the
fusion processing is needed to reduce the redundant performance of data aggregation methods in wireless
information. Another advantage is those reduces the traffic sensor network, Such as energy saving and delay. Data
load and conserve energy of the sensors. Disadvantage: The aggregation is the process, in which aggregating the data
cluster head means data aggregator nodes send fuse these packet coming from the different sources; the number of
data to the base station .this cluster head or aggregator transmission is reduced. With the help of this process we
node may be attacked by malicious attacker. If a cluster can save the energy in the network. Delay is the latency
head is compromised, then the base station (sink) cannot be connected with aggregation data from closer sources may
ensure the correctness of the aggregate data that has been have to held back at intermediate nodes in order to
send to it. Another drawback is existing systems are several combine them with data from source that are farther away.
copies of the aggregate result may be sent to the base Basically aggregation method based on the position of the
station (sink) by uncompromised nodes .It increase the sources in the network, number of sources and the network
power consumed at these nodes. topology. If the examine the factors, we consider the two
models of the source placement. The event radius (ER)
3.2 Performance measure of data aggregation model and random source model [14]. The modelling says
There are very important performance measures of data us that where the source are clustered near each other or
fusion algorithm. These performances are highly located randomly, significant energy gains are possible
dependent on the desired application. with data aggregation. These gains are greatest when the
Energy Efficiency: By the data-aggregation scheme, we can number of sources is large, and when the sources are
increase the functionality of the wireless sensor network. In located relatively close to each other and far from base
which every sensor nodes should have spent the same station. The modelling through, also seems to the suggest
amount of energy in every data gathering round. A data- that aggregation latency could be non negligible.
aggregation scheme is energy efficient if it maximizes the
functionality of the network. Network lifetime, data
accuracy, and latency are some of the significant
performance measures of data-aggregation algorithms. The
4. DATA AGGREGATION APPROACHES IN
definitions of these measures are highly dependent on the WIRELESS SENSOR NETWORK
desired application.
Network lifetime: The network lifetime is defining the Data aggregation process is performed by specific routing
number of data fusion rounds. Till the specified percentage protocol. Our aim is aggregating data to minimize the
of the total nodes dies and the percentage depend on the energy consumption. So sensor nodes should route packets
application .If we talk about some based on the data packet content and choose the next hop
application,simultenously working of the all the sensor in order to promote in network aggregation. Basically
nodes is crucial hence the lifetime of the network is routing protocol is divided by the network structure, thats
number of round until the first nodes which improves the why routing protocols is based on the considered
energy efficiency of nodes and enhance the lifetime of approaches.
whole network.
IJSER 2011
272
ISSN 2229-5518
4.1 Tree-Based Approach BS

The tree based approach is defining aggregation from
constructing an aggregation tree. The form of tree is Sensors Cluster
minimum spanning tree, sink node consider as a root and
source node consider as a leaves. Information flowing of
data start from leaves node up to root means sink(base
station).Disadvantage of this approach, as we know like
wireless sensor network are not free from failure .in case of
data packet loss at any level of tree, the data will be lost not
only for single level but for whole related sub tree as well.
This approach is suitable for designing optimal aggregation
techniques. Madden et al. in [6] data centric protocol know
as Tiny aggregation (TAG) approach. The working of TAG
is depending on two phases: distributed phase and
collection phase.
BS Base station
Level 2
Figure 4 Cluster based sensor network. The arrows indicate wireless
communication links
Level 1 K. Dasgupta et al. in [7] proposed a maximum lifetime

data aggregation (MLDA) algorithm which finds data
gathering schedule provided location of sensors node and
base-station, data packet size, and energy of each sensor
Level 0
node. A data gathering schedule specifies how data packet
are collected from sensors node and transmitted to base
station for each round. A schedule can be thought of as a
Sensor collection of aggregation trees. In [4], they proposed
heuristic-greedy clustering-based MLDA based on MLDA
algorithm. In this they partitioned the network in to cluster
and referred each cluster as super-sensor. They then
Figure 3 Tree based data aggregation in wireless sensor network
compute maximum lifetime schedule for the super-sensors
4.2 Cluster-Based Approach and then use this schedule to construct aggregation trees
In energy-constrained sensor networks of large size, it is for the sensors. W. Choi et al. in [1] present a two-phase
inefficient for sensors to transmit the data directly to the clustering (TPC) scheme. Phase I of this scheme creates
sink In such scenarios, Cluster based approach is clusters with a cluster-head and each node within that
hierarchical approach. In cluster-based approach, whole cluster form a direct connects with cluster-head. Phase I the
network is divided in to several clusters. Each cluster has a cluster-head rotation is localized and is done based on the
cluster-head which is selected among cluster members. remaining energy level of the sensor nodes which minimize
Cluster-heads do the role of aggregator which aggregate time variance of sensors and this lead to energy saving
data received from cluster members locally and then from unnecessary cluster-head rotation. In phase II, each
transmit the result to base station (sink). Recently, several node within the cluster searches for a neighbor closer than
cluster-based network organization and data-aggregation cluster-head which is called data relay point and setup up a
protocols have been proposed for the wireless sensor data relay link. Now the sensor nodes within a cluster
network. Figure 4 shows a cluster-based sensor network either use direct link or data relay link to send their data to
organization. The cluster heads can communicate with the cluster head which is an energy efficient scheme. The data
sink directly via long range transmissions or multi hopping relay point aggregates data at forwarding time to another
through other cluster heads. data relay point or cluster-head. In case of high network
density, TPC phase II will setup unnecessary data relay
link between neighbors as closely deployed sensor will
sense same data and this lead to a waste of energy.
4.3 Multi-path Approach
IJSER 2011
273
ISSN 2229-5518
The drawback of tree based approach is the limited

Directed Diffusion - - -
robustness of the system. To overcome this drawback, a
new approach was proposed by many researchers .in
PEGASIS - - -
which sending partially aggregated data to single parent
node in aggregation tree, a node could send data over
DB-MAC - - -
multiple paths. In which each and every node can send
data packets to its possibly multiple neighbours. Hance
EADAT - - -
data packet flow from source node to the sink node along
multiple path, lot of intermediate node between source
LEACH - -
node to sink node so aggregation done in every -
intermediate node. Using this approach we will make the

system robust but some extra overhead. The example of Cougar - -
this approach like ring topology, where network is divided -
in to concentric circle with defining level levels according
to hop distance from sink.[3]propose a new strategy have Synopsis - -
both issues : energy efficiency and robustness. In which -
single path to connect each node to the base station it is
energy saving but high risk of link failure. But on the other Diffusion - -
-
head multipath approach would require more nodes to
participate with consequent waste of energy. Authors
Tributaries and Deltas - -
present a clever use of multi-path only when there is loss of -
packet which is implemented by smart caching of data at
sensor nodes. Authors also argue that in many practical
situation data may be gathered only from a particular
5. DATA AGGREGATION FUNCTION IN
region, so they use a different approach that relies on a
spanning tree and provides alternative paths only when a WIRELESS SENSOR NETWORK.
malfunctioning is detected. Algorithm adopts a tree-based Many effective type of data aggregation function is needed in
approach for forwarding packets through the network. In wireless sensor network. These functions are closely related
the ideal situation when no failures occur, this is certainly to sensor network application. Such as mean quantile,
the best choice, as the minimum number of nodes is medium, count, average, max, and min.
engaged in the transmission phase. In the presence of link 5.1 Duplicate sensitive and duplicate insensitive
or node failures, the algorithm will discover alternative Aggregation function may be average and minimum. If we
paths, so as ensure the delivery of as many packets as use average function , it take as a duplicate sensitive and
possible within the time constraints. The problem with this minimum function is take as duplicate insensitive function
approach is that it may cause the arising of hot spots and in wireless sensor network. Data aggregated in the network
nodes along preferred paths will consume their energy on that time same data consider multiple times. We can
resources quickly, possibly causing disconnection in the used duplicate function then the final result depends on
network. the number of the times and same value has been
considered otherwise aggregation function is said to be
4.4 Hybrid Approach duplicate insensitive.
Hybrid approach followed between tree, cluster based and 5.2 Lossy and lossless
multipath scheme. In which the data aggregation structure Data packet can be aggregated with the help of lossy
can adjusted according to specific network situation and to aggregation or by lossless aggregation. Lossy aggregation
some performance statistics. approach does not follow a perfect reconstruction but
lossless aggregation ensures a complete recovery of all
Table 1 individual sensor data at base station (sink) [2].
Routing protocol for Tree, cluster, Multipath and Hybrid
approach. 6. DATA REPRESENTATION IN WIRELESS
SENSOR NETWORK.
Protocols/algorithms Tree Cluster Multipath Hybrid Data representation is the effective way to representation
the data. Wireless sensor network is consisting a large

TAG - - - number of small sensor nodes. These are resource
constraint, due to limited resource constraint it needs to
decide whether to store, compress, discard and transmit
IJSER 2011
274
ISSN 2229-5518
data. All this requirement wants a suitable way to security issues is data integrity with the help of integrity
represent the information any type of structure are we reduce the compromised sensor source nodes or
common to all sensor node in the network.[14] aggregator nodes from significantly altering the final
aggregation value. Sensor node in a sensor network is
7. SECURITY ISSUES IN DATA easily to compromised. Compromised nodes have a
AGGREGATION FOR WIRELESS SENSOR capability to modify or discard messages. Method of secure
NETWORK data aggregation: There are two type of method for
There are two type of securities are require for data securing data hop by hop encryption and end to end
aggregation in wireless sensor network, confidentiality and encryption, both methods follows some step.
integrity. The basic security issue is data confidentiality, it 1. Encryption process has to be done by sensing nodes in
is protecting the sensitive data transmission and passive wireless sensor network.
attacks, like eavesdropping. If we talk about hostile 2. Decryption process has to be done by aggregator nodes.
environment so data confidentiality is mainly used because 3. After that aggregator nodes aggregates the result and
wireless channel is vulnerable to eavesdropping by then encrypt the result again.
cryptography method. The complicated encryption and 4. The sink node gets final aggregated result and decrypt it
decryption operations such as modular multiplication. The again.
8. CONCLUSION also discuss the various approaches for data aggregation or
also discuss the advantage and disadvantages and various
In this paper we present wireless sensor network is consist a performance measures of the data aggregation.
large number of sensor node. And these nodes are resource
constraint. Thats why lifetime of the network is limited so the ACKNOWLEDGMENT
various approaches or protocol has been proposed for This work was supported in part by a grant from NIT
increasing the lifetime of the wireless sensor network. In this Hamirpur (himachal Pradesh).
paper we discuss the data aggregation are one of the important
techniques for enhancing the life time of the network. And
[8] M. Ding, X. Cheng and G. Xue, Aggregation Tree

REFERENCES Construction in Sensor Networks, 2003 IEEE 58th Vehic.
Tech. Conf., vol. 4, no. 4, Oct. 2003, pp. 216872.
[1] I. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci,
[9] K. Vaidhyanathan et al., Data Aggregation Techniques in
A Survey On Sensor Networks, IEEE Communications
Sensor Networks, Technical Report, OSU-CISRC-11/04-
Magazine, Volume 40, Number 8, pp.102-114, 2002.
TR60, Ohio State University, 2004
[2] T. Arampatzis , J. Lygeros, and S. Manesis, A Survey Of
[10] V. Raghunathan, C. Schugers, S Park, and M.B. Srivastava,
Applications Of Wireless Sensors And Wireless Sensor
Energy-Aware Wireless Microsensor Networks, IEEE
Networks, In Mediterranean Conference On Control And
Signal Processing Magazine, Volume 19, Number 2, pp. 40-50,
Automation MED05, Nicosia, Cyprus, 2005.
2002.
[3] L. Gatani, G. Lo Re, and M. Ortolani, Robust and Efficient
[11] Lin F.Y.S., Yen H.H., Lin S.P., Wen Y.F.: MAC aware
Data Gathering for Wireless Sensor Networks, in Proceeding
energyefficient data-centric routing in wireless sensor
of the 39th Hawaii International Conference on System Sciences
2006 networks. Proc. IEEE Int. Conf. Commun. (ICC), 2006, vol.
8, pp. 34913496
[4] K. Dasgupta, K. Kalpakis, and P. Namjoshi, An Efficient
[12] Krishnamachari B., Estrin D., Wicker S.: Modeling
Clustering-based Heuristic for Data Gathering and
datacentric routing in wireless sensor networks, USC
Aggregation in Sensor Networks, IEEE 2003
Computer Engineering Technical Report, CENG 02-14, 2002.
[5] E. Fasolo, M. Rossi, J. Widmer, and M. Zorzi, In-Network
[13] H.H. Yen, C.L. Lin: Integrated channel assignment and data
Aggregation Techniques for Wireless Sensor Networks: A
aggregation routing problem in wireless sensor networks,
Survey, IEEE Wireless communication 2007.
[6] S. Madden et al., TAG: a Tiny Aggregation Service for Ad- IET Communications, 2009, Vol. 3, Iss. 5, pp. 784793.
hoc Sensor Networks, OSDI 2002, Boston, MA, Dec. 2002. [14] Ramesh Rajagopalan and Pramod K. Varshney, Syracure
University Data aggregation techniques in sensor network :
[7] K. Dasgupta et al., Maximum Lifetime Data Gathering and
Aggregation in Wireless Sensor Networks, In Proc. of IEEE A survey IEEE communication survey & tutorial.4th
quarter 2006
Networks02 Conference, 2002.
IJSER 2011
275
ISSN 2229-5518
Slant Transformation as a Tool for Pre-

processing in Image Processing
Nagaraj B Patil, V M Viswanatha , Dr. Sanjay Pande MB
Abstract The Slantlet Transform (SLT) is a recently developed multiresolution technique especially well-suited for piecewise linear data.
The Slantlet transform is an orthogonal Discrete Wavelet Transform (DWT) with 2 zero moments and with improved time localization. It also
retains the basic characteristics of the usual filterbank such as octave band characteristic and a scale dilation factor of two. However, the
Slantlet transform is based on the principle of designing different filters for different scales unlike iterated filterbank approaches for the DWT.
In the proposed system, Slantlet transform is implemented and used in Compression and Denoising of various input images. The perfor-
mance of Slantlet Transform in terms of Compression Ratio (CR), Reconstruction Ratio (RR) and Peak-Signal-to-Noise-Ratio (PSNR) present
in the reconstructed images is evaluated. Simulation results are discussed to demonstrate the effectiveness of the proposed method.
Index TermsDiscrete Wavelet transform,Compression ratio, Data Compresion, Peak-signal-to-ratio(PNSR), Coding Inter Pixel, Slantlet
Coefficients, Choppy Images.
1 INTRODUCTION
F or many decades, scientists wanted more appropriate

functions than the sines and cosines, which comprise
shorter length while satisfying orthogonality and zero
moments condition. The basis function retains the octave-
the basis of Fourier analysis, to approximate choppy im- band characteristic. Thus Slantlet transform has been
ages. By their definition, these functions are non-local used as a tool in devising an efficient method for com-
(and stretch out to infinity). Therefore they do a very poor pression and denoising of various images.
job in approximating sharp spikes. Wavelets are functions 2. Literature survey
that satisfy certain mathematical requirements and are G.K. Kharate, A.A. Ghatol, et.al [2], proposed an
used in representing data or other functions. This makes algorithm based on decomposition of images using Dau-
wavelets interesting and useful. But with wavelet analy- bechies wavelet basis. Compression of data is based on
sis, approximating functions can be used that are con- reduction of redundancy and irrelevancy. They have pro-
tained neatly in finite domains. Wavelets are well suited posed an adaptive threshold for quantization, which is
for approximating data with sharp discontinuities [1]. based on the type of wavelet, level of decomposition and
The Discrete Wavelet transform (DWT) is usually nature of data.
carried out by filterbank iteration, but for a fixed number Sos S. Agaian, Khaled Tourshan, et.al [3] propos-
of zero moments it does not yield a discrete time basis es a novel approach for the parameterization of the slant-
that is optimal with respect to time localization. The let transform with the classical slantlet and the Haar
Slantlet transform is an orthogonal DWT with 2 zero transforms are special cases of it is presented. The slantlet
moments and with improved time localization. The Slant- transform matrices are constructed first and then the fil-
let transform has been developed by employing the terbank is derived from them. The parametric slantlet
lengths of the discrete time basis function and their mo- transform performance in image and signal denoising is
ments as the vehicle in such a way that both time- discussed.
localization and smoothness properties are achieved. Us- Panda. G, Dash. P K, et.al [4] proposes a novel
ing Slantlet transform it is possible to design filters of approach for power quality data compression using the
IJSER 2011
276
ISSN 2229-5518
slantlet is presented and its performance in terms of com-

pression ratio (CR), percentage of energy retained and
mean square error present in the reconstructed image is
assessed.
Lakhwinder Kaur, Savita Gupta, et.al [5] pro-
posed an adaptive threshold estimation method for image
denoising in the wavelet domain based on the genera- Fig 3.1. Slantlet filterbank
lized Guassian distribution (GGD) modeling of subband
coefficients. In the Slantlet filterbank, each filter gi(n) does not f
Lei Zhang; Bao, P; Xiaolin Wu [8] proposed a appear together with its time reverse. While hi(n) does not
wavelet-based multiscale linear minimum mean square- appear with its time reverse, it always appears paired
error estimation (LMMSE) scheme for image denoising is with the filter fi(n). In addition, note that the l-scale and (l
proposed, and the determination of the optimal wavelet + 1) scale filterbanks have in common the filters gi(n) for
basis with respect to the proposed scheme is discussed. i=1,., l-1 and their time-reversed versions.
The overcomplete wavelet expansion (OWE), which is The Slantlet filterbank analyzes scale i with the
more effective than the orthogonal wavelet transform filter gi(n) of length 2i+1. The characteristics of the Slantlet
(OWT) in noise reduction, is used. filterbank:
1.Each filterbank is orthogonal. The filters in the synthe-
3.The Slantlet Transform
sis filter bank are obtained by time reversal of the analysis
The Slantlet filterbank is an orthogonal filter filter.
bank for the discrete wavelet transform, where the filters 2.The scale-dilation factor is 2 for each filterbank.
are of shorter support than those of the iterated D2 filter- 3.Each filterbank provides a multiresolution decomposi-
bank tree. This filterbank retains the desirable characteris- tion.
tics of the usual DWT filterbank. 4.The time localization is improved with a degradation of
frequency selectivity.
The Slantlet filter bank shown in Fig 3.1. is generalized as
5.The Slantlet filters are piecewise linear.
follows. The l-scale filter bank has 2l channels. The low-
pass filter is to be called hl(n). The filter adjacent to the

lowpass channel is to be called fl (n). Both hl(n) & fl(n) are
4.Compression
to be followed by down sampling by 2l. The remaining 2l
The term data compression refers to the process
- 2 channels are filtered by gi(n) & its shifted time-reverse
of reducing the amount of data required to represent a
for i =1,., l-1. Each is to be followed by down sampling
given quantity of information. It is useful both for trans-
by 2i+1. x(n) Slantlet filter Coefficients
mission and storage of information[7].
H2(Z) The encoder of the Fig 4.1 is responsible for re-

4
ducing the coding interpixel and/or psychovisual redun-
F2(Z) 4
dancy of the input data. In the first stage of the encoding
IJSER 2011
G1(Z) 4
Z-3 G1(1/Z) 4
277
ISSN 2229-5518
process the mapper transforms the input data into a for- The decoder contains only two components: a
mat designed to reduce the interpixel redundancies. symbol decoder and an inverse mapper. The percentage
of reconstruction ratio can be defined as
Vector norm of the retained SLT Coefficients after
threshold
100
Vector norm of the original SLT coefficients
Compression of data without degradation of data

quality is possible because the data contains a high degree
of redundancy. The higher the redundancy the higher the
achievable compression. The data compression methods

Fig4.1.General Block diagram for compression
can be broadly classified into
1. Lossless compression method

The second stage, or quantizer block, reduces the accura-
2. Lossy compression method.
cy of the mappers output in accordance with a predifined
fidelity criterion- attempting to eliminate only psychovi- 5.DENOISING
sually redundant data. This operation is irreversible and The de-noising objective is to suppress the noise
must be omitted when error-free compression is desired. part of the signal s and to recover f while retaining as
In the third and final stage of process, a symbol coder much as possible the important signal features.
creates a code (that reduces coding redundancy) for the The noisy signal is basically of the form
quantizer output and maps the output in accordance with s(n) = f(n) + e(n), where e(n) is the gaussian white noise
the code. N(0,1) and time n is equally spaced.
If n1 and n2 denotes the number of informations In the recent years there has been a fair amount
carrying units in the original and encoded data respec- of research on wavelet threshold and threshold selection
tively, the compression that is achieved can be quantified for image de-noising, because wavelet provides an ap-
numerically via the compression ratio propriate basis for separating noisy signal from the data
CR = n1/n2 signal. Threshold is a simple non-linear technique, which
To view and/or use a compressed (i.e., encoded) operates on one wavelet coefficient at a time. Replacing
data, it must be fed into a decoder, where a reconstructed the small noisy coefficients by zero and inverse wavelet
output data is generated. Error can be defined between f transform on the result may lead to reconstruction with
and f1 the essential signal characteristics and with less noise. The
e = f1- f ..4.1 Peak-Signal-To-Noise-Ratio(PSNR) for any given data can
IJSER 2011
278
ISSN 2229-5518
be given as, PSNR=20*log10(1256/ (original data-de- If a pixel in the image has intensity less than the
noised data)). threshold value, the corresponding pixel in the resultant

image is set to white. Otherwise, if the pixel is greater
than or equal to the threshold intensity, the resulting pix-
6.Design Procedure
el is set to black. Soft threshold is used for image com-
6.1.Algorithm for signal compression
pression.
The design procedure for signal compression
3.Reconstruction
contains three steps:
Compute wavelet reconstruction based on the
1. Decomposition
modified coefficients. This step uses the inverse Slantlet
The Slantlet transform is used for decomposition transform to perform the wavelet reconstruction.
and applied to each row and then again applied to the 3.1. Find out whether the signal is of finite duration
resulting information in each column. The steps used for (power of 2).
decomposition 3.2. Initialize the moment vectors.
1.Find out whether the input signal is power of 2. 0 = 1:N/2
1.2. Separate the odd and even moment vectors along the 1 = 0
length of the input signal. 3.3. Using DC and linear Slantlet coefficients, compute
0(n;l) and 1(n;l)
1.3. Since the filters are piecewise linear, each filter can be
m = 2^l;
represented as the sum of a DC and a linear term.
0(n;l)=s(1)/(m)*s(2)*(3*(m-1)/(m*(m+ 1)));
1.4. The DC and linear moments at scale i can be com-
1(n;l) = s(2) * (-2 * (3/(m*(m^2 - 1))) 3.4. Then
puted from the DC and linear moments at the next finer
compute 0(n;i) and 1(n;i) for decreasing values of i by
scale (i-1).
updating 0 and 1 using Slantlet coefficients.
The image obtained after 2-D Slantlet transform
6.2. Algorithm for signal De-noising
can be shown as Input image is added with noise. For this noisy
image, apply Slantlet transform. Steps 1 and 3 of the sec-
LL3 LH3 LH2
HL3 HH3 LH1 tion 6.1 remains unchanged. But in the step 2, soft thre-
shold is applied for de-noising. Soft threshold is an exten-
HL2 HH2
sion of hard threshold, first setting to zero the elements
whose absolute values are lower than the threshold, and
HL1 HH1
then shrinking the nonzero coefficients towards 0. The
output of the step3 in section 6.1 is the de-noised version
Figure 5.3. Decomposed image after 2-D Slantlet Trans- of the original input image.
form
7. Results and Analysis

Computer simulation study on Compression and
De-nosing were carried out on some standard images to

2. Threshold the coefficients
IJSER 2011
279
ISSN 2229-5518
the Matlab programs of the Slantlet transform. Some 2D

inputs include images of a cameraman, MRI and Testpart.
Implementation is carried out using Matlab.
7.1. Simulation Results of Compression
The Compression Ratio (CR) and Reconstruction Figure 7.1. Graph for Cameraman
Ratio (RR) for different input images are tabulated as
shown in the table 6.1 and table 6.2. The graphs are plot-
ted against various threshold values and the ratios. 7.2. Simulation results of de-noising
For the various threshold values, the percentage
of signal to noise ratio for different images are tabulated
as shown in the table 6.3 and 6.4. The graphs are plotted
Thr Cameraman MRI Testpart against various threshold values and the PSNR.
value % CR % % % % % RR
RR CR RR CR
0 0 100 0 100 0 100

Table 7.2. Percentage of Signal to noise ratio (PSNR) for dif-
0. 5 14.73 100 52.9 99.9 11.6 100
ferent Images
9 5 8
Thr PSNR PSNR
PSNR (Ca-
5.0
Value 64.68
99.9 70.1 (MRI)
99.9 54.2(Testpart)
99.99
meraman)
9 0 7 5
0.05 19.0813 21.2111 19.0790
10.0 76.02 99.9 79.9 99.8 69.9 99.99
0.5 19.2087 21.0978 19.1190
9 2 7 2 30
5.0 20.0354 21.8754 16.8582
25
100
10.0 98.0220.9197
99.8 98.822.7155
96.6 95.920.6429
99.69
20 Cameraman
PSNR
20.0 0
22.4248 6 24.1644
6 7 21.9417
15 MRI
50.0 99.8123.5620
500 99.2 99.825.2520
89.2 99.622.1960
98.07
10
Testpart
4 1 8 0 5
5000 99.97 96.2 99.9 18.7 99.9 94.74 0
2 9 9 7 0.05 0.5 5 10 20 50
Threshold
Table 7.1. Percentage of Compression Ratio(CR) and
Reconstruction Ratio(RR) for different Images
Figure 7.2. Graph between Threshold and PSNR for different

Images.
120 8. Snapshots
100
IJSER 2011
80
% Ratio
60
%Cr
40 % RR
20
0
280
ISSN 2229-5518
The snapshots of the implementation of Compres-

sion & De-noising for different images are shown
Figure 8.3. (a) Input image(testpart) (b) Decomposed image

(c) Compressed image (d) Reconstructed image
Figure 8.1. (a) Input image(cameraman) (b) Decomposed

image (c) Compressed image (d) Reconstructed image.
Figure 8.4. (a) Input image(cameraman) (b) Noisy image (c)

De-noised image .
Figure 8.2. (a) Input image(MRI) (b) Decomposed image (c)
Compressed image
(d) Reconstructed image.
Figure 8.5. (a) Input image(MRI) (b) Noisy image (c)

Denoised image.
IJSER 2011
281
ISSN 2229-5518
Authors thank Singhania University Rajastan and Vidya

Vikas Institute of Technology, Mysore.
Bibliography
1. I. W. Selesnick, The Slantlet Transform, IEEE
Transaction on signal processing, vol 47, No. 5,
May 1999.
2. David l. Donoho Smooth wavelet decomposi-
Figure 8.6. (a) Input image(Testpart) (b) Noisy image (c) De- tions with blocky coefficient kernels May, 1993.
noised image. 3. Vaidyanathan, p. p., Multirate systems and filter
9.Conclusion banks, prentice hall inc., Englewood cliffs, new
The Slantlet transform orthogonal filterbank uses jersey, 1993.
the filters of shorter length as compared to iterated 2- 4. M. Lmg, H. Guo, et. al Noise reduction using an
scale DWT filterbank. The number of filters increases
undecimated discrete wavelet transform, IEEE
according to the order. The Slantlet transform gives better
signal processing , vol 3, no. 1, Jan 1996.
compression result for the piecewise linear data. The
5. Aali M. Reza ,From Fourier transform to Wave-
Slantlet transform has been applied for compression and
let transform, October 27, 1999 white paper.
de-noising of various images. The matlab programs for
6. Paolo Zatelli, Andrea Antonello, New Grass
the Slantlet transform applied for compression and de-
noising are developed and the computer simulation is modules for multiresolution analysis with wave-
carried out for some test images. It is observed that as the lets, proceedings of the open source gis - grass
threshold level increases better compression ratio and users conference, September 2002.
PSNR can be achieved for the test data. 7. Kenneth Lee, Sinhye Park, et.al, Wavelet-based
image and video compression, Tcom 502, April,

10. Future scope 1997.
The present work can be extended to compare 8. Dr. Guangyi Chen , Signal or Image denoising
the performance of applying Wavelet transform and using wavelets, multiwavelets and curvelets,
Slantlet transform on images and also study the characte- 2004.
ristics/performance benefits of Slantlet transform over
Wavelet transform. This work can be enhanced to wavelet

packets, which offers a more complex and flexible analy-
sis, because the details as well as the approximations are
split.
Acknoweledgements
IJSER 2011
282
ISSN 2229-5518
Separation of Concerns in VoiceXML

Applications
Sukhada P. Bhingarkar
Abstract Many commercial applications provide customer services over the web like flight tracking, emergency notification, order
inquiry etc. VoiceXML is an enabling technology for creating streamlined speech-based interface for such web-based information ser-
vices. Whereas in computing, aspect-oriented programming (AOP) is a programming paradigm, which aims to increase modularity.
AOP includes programming methods and tools that support the modularization of concerns at the level of the source code. The aim of
this paper is to integrate AOP with VoiceXML. Aspect-Oriented Programming (AOP) encapsulates common low-level scattered code
within reusable components called aspects. There are certain tags in VoiceXML like <nomatch>, <noinput>, <error> which appear
commonly in every VoiceXML document. These tags can be considered as the concerns and can be put inside an aspect. This elimi-
nates the need to programmatically write these tags in every VoiceXML document and modularizes the crosscutting-concerns.
Index Terms AOP, AspectJ, ASR, IVR, TTS, VoiceXML, VXML
1 INTRODUCTION
T he WWW has become primary source of information

all over the world and accounts for the major propor-
tion of entire Internet traffic. The next leading edge
tries to integrate this feature of AOP with VoiceXML.
VoiceXML has certain tags which appear across the scope
of every VoiceXML page. These tags are crosscutting
for the research on the web is to make it accessible via concerns that can be defined in a modular fashion with
voice and audio. Considerable work has been done in this
the help of aspect-oriented programming. This helps
direction, which includes the design of VoiceXML and
voice browsers. It allows voice applications to be devel- modularization and code reuse in VoiceXML based
oped and deployed in an analogous way to HTML for applications.
visual applications.
Aspect orientation is not a completely new approach to The rest of the paper is organized in following way:
writing software. Aspect orientation is becoming a com- section II presents related work in this field. Section III
and IV discusses VoiceXML and AOP respectgively. The
monly adopted and de facto approach to practicing older
paper, then, proposes an integration of AOP with
ideas that can be traced to almost the beginning of soft- VoiceXML in Section V. Finally, the paper finishes with
ware development. Development environments and tools conclusion in Section VI.
that weave code, pragma instructions, and even debug-
gers all contain some of the behavior that underlies the
2 RELATED WORK
aspect-oriented approach. It is a more modular imple-
mentation of the advantages that these technologies have During past few years, several applications are
developed to assist visually impaired, technologically
brought to their own domains in the past. Aspect-
uneducated and underprivileged people to acoustically
oriented programming entails breaking down program
access the information that was originally intended to be
logic into distinct parts called as concerns. All accessed visually via a personal computer (PC). Voice
programming paradigms support some level of grouping response facilities are used for various kinds of
and encapsulation of concerns into separate, independent information over the phone: time, weather, horoscopes,
entities by providing abstractions (e.g., procedures, sports, cultural events and so on. Nuance is one of the
modules, classes, methods) that can be used for providers for speech and imaging solutions for businesses
implementing, abstracting and composing these concerns. and customers around the world that is based on IVR
But some concerns defy these forms of implementation systems. The VoiceXML along with Nuance 8.5 speech
recognition software can reduce cost and effort of
and are called crosscutting concerns because they "cut
deploying voice-driven services. In [1], a VoiceXML-
across" multiple abstractions in a program. This paper
driven audio wiki application is presented that is
IJSER 2011
283
ISSN 2229-5518
accessible via both the Public Switched Telephone Automatic Speech Recognition (ASR) and/or touchtone
Network and the Internet. Users may access wiki content (DTMF keypad) for input, and prerecorded audio and
via fixed or mobile phones or via a PC using web browser text-to-speech (TTS) synthesis for output. Numerous
or a Voice over IP service. Silog is a biometric authentica- commercial vendors such as IBM, TellMe and BeVocal
tion system that extends the conventional PC logon provide voice browsers that can be used to play Voice-
process using voice verification [2]. SeeCCT is a XML documents. Current voice interfaces to the web are
prototype of a multimodal social networking system of two types: voice interface to screen display and voice-
designed for sharing geographical bookmarks [3]. In [4], only interface. The dynamic voice interface presented in
VoiceXML portal is developed that allows people with a this paper is a voice-only interface that uses a telephone
mobile or usual phone to get informed about cultural as an input and output device. Figure 1 shows the core
activities by dialing a phone number and by interacting architecture of VoiceXML applications.
with a computer via voice. Domain-specific dialogs are
created in native languages viz. slavic languages using
VoiceXML [5]. HearSay is a non-visual Web browser
developed for visually impaired users [6].
AspectJ is one of the oldest and well-known aspect lan-
guages and it helped to bring AOP to the mainstream.
AspectJ is an extension of the Java language which de-
fines a special syntax for declaring aspects. The first ver-
sions of AspectJ featured compile-time source code weav-
ing and bytecode weaving. It later merged with Aspect-
Werkz which brought load-time weaving as well as As- Fig. 1. Architectural Model of VoiceXML
pectWerkzs annotation style to the language. AOP is
used in a variety of fields. Aspect-oriented software de- A document server processes requests from a client ap-
velopment had played an important role in the design plication, the VoiceXML Interpreter, through the Voice-
and implementation of PUMA which is a framework for XML interpreter context. The server produces VoiceXML
the development of applications that analyze and trans- documents in reply, which are processed by the Voice-
form C or C++ source code [7]. A framework for middle- XML Interpreter. The VoiceXML interpreter context may
ware design is invented which is based on the Concurrent monitor user inputs in parallel with the VoiceXML inter-
Event-based Aspect-Oriented paradigm [8]. A fully dy- preter. The implementation platform is controlled by the
namic and reconfigurable monitoring system is designed VoiceXML interpreter context and by the VoiceXML in-
based on the concept of Adaptable Aspect-Oriented Pro- terpreter.
gramming (AAOP) in which a set of AOP aspects is used
to run an application in a manner specified by the adap-
tability strategy [9]. The model can be used to implement
4 ASPECT ORIENTED PROGRAMMING (AOP)
systems that are able to monitor an application and its The main idea of AOP is to isolate the cross-cutting con-
execution environment and perform actions such as
cerns from the application code thereby modularizing
changing the current set of resource management con-
them as a different entity. A cross-cutting concern is be-
straints applied to an application if the applica-
tion/environment conditions change. An aspect-oriented havior, and often data, that is used across the scope of a
approach is advocated as an improvement to the object piece of software. It may be a constraint that is a characte-
oriented approach in dealing with the issues of code tan- ristic of your software or simply behavior that every class
gling and scattering in case of multilevel security [10]. must perform. The most common example of a cross-
cutting concern is that of logging. Logging is a cross-
3 VOICEXML cutting concern because it affects many areas across the
software system and it intrudes on the business logic.
VoiceXML is HTML of Voice Web. It is W3Cs standard
XML format for specifying interactive voice dialogues Logging is potentially applied across many classes, and it
between a human and a computer. VoiceXML is a mar- is this form of horizontal application of the logging aspect
kup language for creating voice user interfaces. It uses that gives cross-cutting its name. Some central AOP con-
cepts are as follows:

Sukhada Bhingarkar is with the Computer Engineering Dept., MIT colle- Aspects: A modularization of a concern that cuts across
gee of Engineering, Kothrud, Pune, India 411038
E-mail:[email protected]
multiple objects. Transaction management is a good
example of a crosscutting concern in J2EE applications.
IJSER 2011
284
ISSN 2229-5518
Join Point: A point during the execution of a program,

such as the execution of a method or the handling of an
exception 5 INTEGRATING AOP WITH VOICEXML
Advice: The code that is executed when an aspect is in- VoiceXML tags such as <nomatch>, <noinput> <er-
ror> are usually common tags in most of the VoiceXML
voked is called advice. Advice contains its own set of rules
documents. Figure 4 shows VoiceXML example that illu-
as to when it is to be invoked in relation to the join point strates basic capabilities of VoiceXML. This VoiceXML is
that has been triggered. Different types of advice include logically divided into three parts: header, body and foo-
"around," "before" and "after" advice. ter. The header part can contain tags showing xml and
vxml version. The body part contains core business logic
Pointcut: A predicate that matches join points. Advice is while footer part has error handing and <naomatch>,
associated with a pointcut expression and runs at any join <noinput> tags.
point matched by the pointcut (for example, the execution
of a method with a certain name).
Figure 2 shows the relationships between join points, as-
pects, pointcuts, advice, and your application classes.
Fig.2. Relationship of AOP concepts with application class [11]
Figure 3 shows a simple business logic class along with Fig. 4. VoiceXML example
an aspect applied to this class.
The commonly occurring tags in VoiceXML can be consi-
dered as crosscutting concerns and are encapsulated in a
special class i.e. an aspect. Figure 5 shows the Servlet that
is used to construct VoiceXML.
Fig. 5. Servlet generating VoiceXML code
doGet() or doPost() methods of Servlet concentrates only

on core logic code i.e. body part of VoiceXML. The header
and footer parts of VoiceXML will be taken care by an
aspect class shown in figure 6.
Fig.3. Business logic class along with aspect
IJSER 2011
285
ISSN 2229-5518
Aspect-Oriented Software Development, March 2010.

[8] Edgar Marques, Lus Veiga, Paulo Ferreira, An extensible
framework for middleware design based on concurrent event-
based AOP, ARM '10: Proceedings of the 9th International
Workshop on Adaptive and Reflective Middleware, November
2010
[9] Arkadiusz Janik, Krzysztof Zielinski, AAOP-based dynamical-
ly reconfigurable monitoring system, Information and Soft-
ware Technology, Volume 52 Issue 4, April 2010
[10] S. Kotrappa, Prakash J. Kulkarni, Multilevel Security Using
Aspect Oriented Programming AspectJ, ARTCOM '10: Pro-
ceedings of the 2010 International Conference on Advances in
Recent Technologies in Communication and Computing, Octo-
Fig. 6. Aspect applied to Servlet ber 2010
[11] Russell Miles, AspectJ Cookbook, O'Reilly, 2004
6 CONCLUSION
VoiceXML is a language to create voice-user interfaces
while AOP allows us to dynamically modify our static
model to include the code required to fulfill the second-
ary requirements without having to modify the original
static model. Integration of AOP with VoiceXML better
separates the concerns of Voice based applications, there-
by providing modularization. This helps developers to
concentrate on core logic and promotes code reuse.
REFERENCES
[1] Constantinos Kolias et al, Design and implementation of a
VoiceXML-driven wiki application for assistive environments
on the web, Personal and Ubiquitous Computing, 2010,
Volume 14, Number 6, 527-539, Springer-Verlag London Li-
mited 2010
[2] Sergio Grau, Tony Allen, Nasser Sherkat, Silog: Speech input
logon, Knowledge-Based Systems, Volume 22, Issue 7, October
2009, Pages 535-539
[3] Stan Kurkovsky et al., Mobile Voice Access in Social Network-

ing Systems, in the Proceedings of 5th IEEE International Con-
ference on Information Technology: New Generations, Las Ve-
gas, USA, 7-9 Apr. 2008
[4] Evangelia Boufardea et al, A Dynamic Voice Portal for Deli-
very of Cultural Content in the Proceedings of 3rd Internation-
al Conference on Internet and Web Applications and Services
(ICIW08), Athens, 8-13 Jun. 2008
[5] Brkic M., Matetic M., VoiceXML for Slavic Languages Applica-
tion Development, in the Proceedings of IEEE international
conference on Human System Interactions, Krakow, Poland, 25-
27 May 2008
[6] Borodin Y, Mahmud J, Ramakrishman IV, Stent A (2007) The
hearsay non-visual web browser. In: ACM international confe-
rence proceeding series, proceedings of the 2007 international
cross-disciplinary conference on web accessibility (W4A), vol
225. Banff, Canada, pp 128129
[7] Matthias Urban, Daniel Lohmann, Olaf Spinczyk, The Aspect-
Oriented Design of the PUMA C/C++ Parser Framework. In:
AOSD '10: Proceedings of the 9th International Conference on
IJSER 2011
286
ISSN 2229-5518
Error of Approximation in Case of

Definite Integrals
Rajesh Kumar Sinha, Satya Narayan Mahto, Dhananjay Sharan
Abstract This paper proposes a method for computation of error of approximation involved in case of evaluation of integrals
of single variable. The error associated with a quadrature rule provides information with a difference of approximation. In
numerical integration, approximation of any function is done through polynomial of suitable degree but shows a difference in
their integration integrated over some interval. Such difference is error of approximation. Sometime, it is difficult to evaluate the
integral by analytical methods Numerical Integration or Numerical Quadrature can be an alternative approach to solve such
problems. As in other numerical techniques, it often results in approximate solution. The Integration can be performed on a
continuous function on set of data.
Index Terms Quadrature rule, Simpsons rule, Chebyshev polynomials, approximation, interpolation, error.
1 INTRODUCTION
T O evaluate the definite integral of a function that

has no explicit antiderivative of whose antideriva-
tive is not easy to obtain; the basic method in-
tion for a given function, concentrating mainly on po-
lynomial approximation. For approximation, there is
considered a polynomial of first degree such as
volved in approximating is numerical quadrature [1]-
y a bx a good approximation to a given function
[4].
for the interval (a, b).
b
f x dx
a
(1)
2 PROPOSED METHOD
n b
2.1 Reflection on Approximation
i.e. i
f ( x) to approximate f ( x)dx . The methodolo-
i 0 a This section cover types of approximation following
gy for computing the antiderivative at a given point, the condition best approximation for a given func-
the polynomial p( x) approximating the function f ( x) tion, concentrating mainly on polynomial approxima-
generally oscillates about the function. This means that tion. In this for approximation, there is considered a
if y p( x) over estimates the function y p( x) in one polynomial of first degree such as y a bx ; a good
interval then it would underestimate it in the next in- approximation to a given continuous function for the
terval [5]. As a result, while the area is overestimated interval (0, 1).
in one interval, it may be underestimated in the next Under the assumption of given concept two following
interval so that the overall effect of error in the two statements may be considered as,
intervals will be equal to the sum of their moduli, in- The Taylor polynomial at x 0 (assuming f (0) exists)
stead the effect of the error in one interval will be neu-
tralized to some extent by the error is the next interval. y f (0) xf (0) (2)
Therefore, the estimated error in an integration formu-
la may be unrealistically too high. In view to above The interpolating polynomial constructed at x 0
discussed facts, the paper would reveal types of ap- and x 1 .
proximation following the condition best approxima-

y f (0) x[ f (1) f (0)] (3)
Rajesh Kumar Sinha is with the Department of Mathematics, NIT Patna,
India. E-mail: [email protected] A justification may be laid that a Taylor or interpolat-
Satya Narayan Mahto is with the Department of Mathematics, M. G. Col-
lege, LNMU, Darbhanga, India.
ing polynomial constructed at some other point would
Dhananjay Sharan is research scholar in the Department of Mathematics, be more suitable. However, these approximations are
NIT Patna, India. designed to initiate the behavior of f at only one or two
points.
IJSER 2011
287
ISSN 2229-5518
Since, the polynomial of first degree in x as shown imum modulus of f ( x) p( x) .

above y a bx follows a good approximation to f To make an evident proof for a given statement an as-
throughout the interval (0,1). Now, for values of a and sumption is made
b , the required mathematical exists such as
f ( x) (a bx) is minimized over all choices of
max f ( x) p( x) 102 Tn ( x) (4)
0 x 1
two values a and b . This expression is said as mini-

max (or Chebyshev) approximation. Instead of mini- Where Tn ( x) is a polynomial of degree n in x with
mizing the minimum error between the (continuous) leading term 2n 1 xn for n 0 . The Tn are known as
function f and the approximating straight-line, the Chebyshev polynomials, the Russian mathematician P.
process of maximizing sum of the moduli of the er- L. Chebyshev (1821 -1894) who has contributed the
rors may be undertaken. equation (4) for the difference between of the first ap-
1
proximation and the polynomial. Now, to determine
For values of a and b , f ( x) ( a bx)dx is mini-
0
the turning value, of Tn the derivative
mized that is called a base L1 approximation. It should
be noted that the L1 approximation provides equal d d d n sin n
T ( x) cos n (5)
weight to all the errors, while the minimax approxima- dx n d dx sin
tion approximate in the error of largest modulus.
Since
Again, stressing on other approximation f which in a
d 1 1
sense, lies between the extremes of L1 and minimax (6)
dx dx sin
approximation. Also, for a fixed value of p 1 , two d
values a and b are formed so that n sin n sin n
1 i.e. Tn( x) n2 (7)
p sin n sin
f ( x) (a bx) dx is minimized and therefore would
0
be suggested as best LP approximation. The above sin

but Lt 1 (8)
maximized expression followed that due to the pres- 0

ence of the P th power, the error of largest modulus
tends to dominate as p increases with f continuous. It Thus Tn(1) n2 (9)
can be shown that, as p the best LP approxima-
tion tends to the minimax approximation which is It can be shown that this is the maximum modulus of
therefore sometimes called the best L approximation. Tn on (-1, 1). If n 10 , say, the maximum modulus of
Thus the LP approximations consist of a spectrum f p 10 2 on (-1, 1) whereas f p 1 . Furthermore,
ranging from the L1 to the minimax approximations. the consideration of an approximation f and the po-
Further, for 1 p , L2 approximation, is the only lynomial pn that interpolates the approximation f at
commonly used and is better known as the best square distinct point x1 ............xn . If such happens then there
approximation.
exists a number x (depending on x ) in certain inter-
val (a, b) such that
2.2 Generalized case for approximation under
certain interval
f ( n 1) ( x )
This section would reveal for giving light on consider- f ( x) pn ( x) ( x x0 )( x x1 )..........( x xn ) (10)
ation of methods of approximating to f , given the (n 1)
values of f ( x) at certain points. If p is same polynomi-
al, approximation to f an application of p would exist f ( n 1) ( x )
f ( x) pn ( x ) n 1 ( x) (11)
an approximation to f . However, there is need to be (n 1)
careful; the maximum modulus of f ( x) p( x) on a where n 1 ( x) ( x x0 )( x x1 ).........( x xn ) .
given interval (a, b) can be much larger than the max-
IJSER 2011
288
ISSN 2229-5518
Thus equation (11) is known as error that exists in fol- Differentiating both sides with respect to x ,
lowing statement. If (a, b) is any interval and also con-
tains (n +1) points x, x0 , x1 ,.......xn . Suppose further it is ds d
pn ( x) p ( x sh) (19)
assumed that f , f .......... f n
exist and are continuous dx ds n 0
on the interval (a, b) and f

n 1
exists for a x b then
1 d s s 2 s n
the error holds as shown in equation (11). Thus, f
n 1 pn ( x)
ex- f 0 f 0 f 0 .... f 0
dx ds 1 2 n
ists on some interval (a, b) which is x , x0 , x1 , x2 ,...., xn . ds
Here, the number x ( a , b) (depending on x). Diffe- (20)
rentiating (11) with respect to x , 1 1 d s
pn ( x) f 0 (2s 1) 2 f 0 ..... n f 0 (21)
h 2 ds n
f ( n 1) ( x) n 1 ( x) d ( n 1)
f ( x) pn ( x ) n 1 ( x) f ( x ) (12)
(n 1) (n 1) dx To calculate n 1 ,
In general, there is nothing further to state about the n 1 ( x) ( x xr ) ( x xr ) (22)

j r
second term on the right of equation (12). We can not
perform the differentiation with respect to x where n 1 would be obtained by means of differen-
( n 1)
of f ( x ) , since x is an unknown function of x . tiating equations (22) such that
Thus, for Integral values of x given in equation (12) is
unless for determining the accuracy with which pn d d
n1 (x) (x xr ) (x x j ) (x xr ) (x xr )
approximately to f . However, if we restrict x to one dx jr dx j r
of the values x0 , x1 ,.......xn , then n 1 ( x) 0 and the un- (23)
known second term on the right of equation (12) be- By putting x xr , the second term on the right of equa-
comes zero. tion (22) becomes zero.
f ( n 1) ( r ) Thus n1 ( x) ( x x j ) (1) n r h n r !(n r )! (24)

jr
f ( xr ) pn ( x) n 1 ( xr ) (13)
(n 1)
Since
x r x j ( r j )h (25)
Where r has been considered when x xr . Now,
applying forward difference to express the polynomial
From (13)
pn ( x) in given form such that
r !(n r )! ( n 1)
f ( xr ) pn ( xr ) (1)( n r ) hn f ( r ) (26)
(n r )!
pn ( x) pn ( x0 sh) (14)
Furthermore, investigating the case of polynomial, an
s s s
pn ( x) f 0 f 0 2 f 0 ....... n f 0 (15) interpretation hold that polynomials are sufficiently
1 2 n accurate for many approximation and interpolation
tasks.
Since x x0 sh (16)
2.3 Degree of Accuracy

dx
i.e. h (17) Now the degree of accuracy of quadrature formula is
ds
the largest positive integer n such that the formula is
Also exact for x k , for each k 0,1,....n . The Trapezoidal and
Simpsons rule have degrees of precision one and
pn ( x) pn ( x0 sh) (18) three, respectively. Integration and summation are li-
near operations; that is,
IJSER 2011
289
ISSN 2229-5518
b b b physical sciences.
( f ( x) g( x))dx f ( x)dx g( x)dx
a a a
(27)
REFERENCES
n n n [1] K. E. Atkinson, An Introduction to Numerical Analysis, Wiley,
NewYork, 1993.
( f ( x ) g( x )) f ( x ) g( x )
i 0
i i
i 0
i
i0
i
(28)
[2] R. E. Beard, Some notes on approximate product integration,
J. Inst. Actur., vol. 73, pp. 356-416, 1947.
[3] C. T. H. Baker, On the nature of certain quadrature formulas
For each pair of integral functions f and g and each and their errors, SIAM. J. Numer anal., vol. 5, pp. 783-804, 1968.
pair of real constants. This implies that the degree of [4] P. J. Daniell, Remainders in interpolation and quadrature for-
mulae, Math. Gaz., Vol. 24, pp. 238-244, 1940.
precision of a quadrature formula in n if and only if [5] R. K. Sinha, Estimating error involved in case of Evaluation of
the error E( p( x)) 0 for all polynomials p( x) of de- Integrals of single variable, Int J. Comp. Tech. Appl., Vol. 2, No.
2, pp. 345-348, 2011.
gree k 0,1,....n , but E( p( x)) 0 for some polynomial [6] T. J. Akai, Applied Numerical Methods for Engineerrs, Wiley,
NewYok, 1993.
p( x) of degree n 1 . [7] L. M. Delves, The Numerical Evaluation of Principal Value
Integrals, Computer Journal, Vol. 10, pp. 389, 1968.
[8] Brain Bradi, A Friendly Introduction to Numerical Analysis,
3 CONCLUSION pp. 441-532, Pearson Education, 2009.
[9] R. K. Sinha, Numerical Method for evaluating the Integrable
Increasing, the degree of the approximating polynomi- function on a finite interval, Int. J. of Engineering Science and
al dose not guarantees better accuracy. In a higher de- Technology, Vol. 2, No. 6, pp. 2200-2206, 2010.
gree polynomial, the coefficients also get bigger which [10] C. E. Froberg, Introduction to Numerical Analysis, Addison-
Wesley Pub. Co. Inc.
may magnify the errors. Similarly, reducing the size of
[11] Ibid, The Numerical Evaluation of class Integrals, Proc. Comb.
the sub-interval by increasing their number may also Phil. Soc.52.
lead to accumulation of rounding errors. Therefore a [12] P .J .Davis and P. Rebinowitz, Method of Numerical Integra-
balance should be kept between the two, i.e. degree of tion, 2nd edition, Academic Press, New York, 1984.
polynomial and total number of intervals. These are
the primary motivations for studying the techniques of
numerical integration/quadrature [6]-[9]. In case of
Simpsons rule technique individually to the subinter-
vals [a ,(a b) 2] and [( a b) 2 , b] ; use error estimation
procedure to determine if the approximation to the
integral on subinterval is within a tolerance of 2 . If
so, then sum the approximations to procedure an ap-
proximation of function f x over interval a, b
within the tolerance . If the approximation on one of
the subintervals fails to be within the tolerance 2 ,
then that subinterval is itself subdivided, and the pro-
cedure is reapplied to the two subintervals to deter-
mine if the approximation on each subinterval is accu-
rate to within 4 . This halving procedure is continued
until each portion is within the required tolerance.
Thus, Numerical analysis is the study of algorithms
that use numerical approximation for the problems of
continuous functions [10]-[12]. Numerical analysis con-
tinues this long tradition of practical mathematical cal-
culations. Much like the Babylonian approximation,
modern numerical analysis does not seek exact an-
swers, since the exact answers are often impossible to
obtain in practice. Instead, much of numerical analysis
is concerned with obtaining approximate solutions
while maintaining reasonable bounds on errors. It
finds applications in all fields of engineering and the
IJSER 2011
290
ISSN 2229-5518
An Assessment model for Intelligence

Competencies of Accounting Information
Systems
Mehdi Ghazanfari, Mostafa Jafari, Saeed Rouhani
Abstract Accounting Information Systems (AIS) as computer-based systems that processes financial information and
supports decision tasks have been implemented in most organizations but, but they still encounter a lack of Intelligence in their
decision-making processes. Models and methods to evaluate and assess the Intelligence-level of Accounting Information
Systems can be useful in deploying suitable business intelligence (BI) services. This paper discusses BI Assessment criteria,
fundamental structure and factors used in the Assessment model. Factors and the proposed model can assess the intelligence
of Accounting Information Systems to achieve enhanced decision support in organizations. The statistical analysis identified five
factors of the Assessment model. This model helps organizations to design, buy and implement Accounting Information
Systems for better decision support. The study also provides criteria to help organizations and software vendors implement AIS
from decision support perspectives.
Index Terms Business Intelligence; Decision Support; Accounting Information Systems; Assessment Model.
1 INTRODUCTION
I nformation and knowledge represent the fundamental

wealth of an organization. Enterprises try to utilize this
wealth to gain competitive advantage when making
performance measures, but before their effort, the mea-
surement and the evaluation in the BI field were re-
stricted to proving the worth of BI investment, and the
important decisions. Enterprise systems like Accounting value of BI. Elbashir et al. (2008) have discussed measur-
Information Systems (AIS) converts and store the data. ing the effects of BI systems on the business process, and
Therefore, it is important to integrate decision-support have presented effective methods of measurement. Lin et
into the environment of these systems. Business intelli- al. (2009) have also developed a performance evaluation
gence (BI) can be embedded in these enterprise systems to model for BI systems using ANP, but they have also
obtain this competitive advantage. treated BI as a separate system.
Today, approaches using an individual system for de- ANP, but they have also treated BI as a separate sys-
cision-support, such as decision-support systems (DSS), tem.
have been replaced by a new, environmental approach. In Organizations usually utilize functional and non-
the past, DSS were independent, separate systems in an functional requirements to assess and select enterprise
organization (island systems). However, enterprise sys- systems like AIS, so the consideration of their decision-
tems are now the foundation of an organization, and support environment as a non-functional requirement,
practitioners are designing BI as an umbrella concept to raises the following questions.
create a decision-support environment for management 1. Which criteria are suitable and effective in the Intel-
(Alter 2004). The increasing trend to use intelligent tools ligence assessment of Accounting Information Systems
in business systems has increased the need for Intelli- (AIS)?
gence Assessment of Accounting Information Systems 2. What is the fundamental structure of these criteria?
(AIS).
There have been some limited efforts to assess BI, but This research was carried out to find answers to the
they have always considered BI as a system that is iso- above questions and to provide a model for efficient deci-
lated from the enterprise systems like AIS. Taking a glob- sion-support by evaluating systems and making BI an
al view, Lnnqvist and Pirttimki (2006) have designed BI integral part of these systems. The rest of this paper is
organized as follows. Section 2, describes brief literature
about AIS. Section 3 is about attempts in previous studies
Mehdi Ghazanfari is currently full professor in Iran University of Science to define Business intelligence (BI). A wide-ranging litera-
and Technology, Tehran, Iran, E-mail: [email protected] ture review about BI and decision-support criteria to as-
Mostafa Jafari is currently assistant professor in Iran University of Science
and Technology, Tehran, Iran sess systems is also summarized in Section 3. Research
Saeed Rouhani is currently PhD Candidate in Iran University of Science methodology and research stages are discussed in section
and Technology, Tehran, Iran, E-mail: [email protected], Con- 4. Section 5 discusses the design of the questionnaire, data
tact:+989122034980
collection, reliability analysis, factor extraction, and label-
IJSER 2011
291
ISSN 2229-5518
ling and assessment model. Finally, Section 6 concludes agents, was first described by McCarthy (1979, 1982) and
the research work, its findings and proposed future re- later developed by an exclusive group of researchers (Da-
search. vid et al., 2002). Extensions to resources, events and
agents include locations (Denna et al., 1993), tasks and
commitments (Geerts and McCarthy, 2002).
2 ACCOUNTING INFORMATION SYSTEMS
An unshakable stream of research exists within the AIS
An Accounting Information Systems (AIS) is defined as a literature that investigates behavioral issues in relation to
computer-based system that processes financial informa- accounting information systems (Sutton and Arnold,
tion and supports decision tasks in the context of coordi- 2002). This stream of research investigates the impact of
nation and control of organizational activities. Extant ac- IT on individuals, organizations and society.
counting information systems research has evolved from An example of behavioral AIS research is a study car-
the source disciplines of Computer science, organization- ried out by Arnold et al. (2004) on the use and effect of
al theory and cognitive psychology. An advantage of this intelligent decision aids. The authors find that smart ma-
evolution is a diverse and rich literature with the poten- chines must be operated by smart people. If users are in-
tial for exploring many different interrelationships among experienced, they will be negatively impacted by the sys-
technical, organizational and individual aspects of judg- tem. Furthermore, they will not learn by experience. Ab-
ment and decision performance. AIS research also spans ernethy and Vagnoni (2004) found that top management
from the macro to the micro aspects of the information uses the newly implemented system for monitoring. Use
system (Birnberg & Shields, 1989; Gelinas et al., 2005). of AIS is found to have a positive effect on cost con-
The comparative advantage of accounting researchers sciousness, but the cost consciousness is hampered if
within the study of IT lies in their institutional accounting people have informal power. In this context, power is an
knowledge. Systems researchers can contribute insights explanatory variable of AIS use.
into the development of systems utilizing technology, and
the other sub-areas can contribute insights into the task
characteristics in the environment. For instance, systems 3 BUSINESS INTELLIGENCE
researchers have extensively investigated group decision Business Intelligence or BI is a grand, umbrella term in-
support systems (GDSS), but they have only recently been troduced by Howard Dresner of the Gartner Group in
considered in auditing. On the other hand, auditing re- 1989 to describe a set of concepts and methods to improve
search has extensively investigated the role of knowledge business decision-making by using fact-based compute-
and expertise. The merging of the two sets of findings rized support systems (Nylund, 1999). The first scientific
may be relevant to AIS design, training, and use. definition, by Ghoshal and Kim (1986) referred BI to a
As comparable term, Management accounting systems management philosophy and tool that helps organiza-
(MAS) also are formal systems that provide information tions to manage and refine business information for the
from the internal and external environment to managers purpose of making effective decisions.
(Bouwens & Abernethy, 2000). They include reports, per- BI was considered to be an instrument of analysis,
formance measurement systems, computerized informa- providing automated decision-making about business
tion systems, such as executive information systems or conditions, sales, customer demand, and product prefe-
management information systems, and also planning, rence and so on. It uses huge-database (data-warehouse)
budgeting, and forecasting processes required to prepare analysis, as well as mathematical, statistical, artificial in-
and review management accounting information. telligence, data mining and on-line analysis processing
Research on management accounting and integrated (OLAP) (Berson and Smith, 1997). Eckerson (2005) un-
information systems (IIS) has evolved across a number of derstood that BI must be able to provide the following
different research streams. Some research streams put tools: production reporting tools, end-user query and re-
heavier emphasis on the management accounting side, porting tools, OLAP, dashboard/screen tools, data min-
while other research streams put emphasis on the infor- ing tools and planning and modelling tools.
mation systems side. Likewise, different research streams BI includes a set of concepts, methods, and processes
approach the topic from different perspectives (Anders to improve business decisions, which use information
Rom & Carsten Rohde, 2007). from multiple sources and apply past experience to de-
A major stream of research within AIS research deals velop an exact understanding of business dynamics (Ma-
with the modeling of accounting information systems. ria, 2005). It integrates the analysis of data with decision-
Several modeling techniques stay alive within the infor- analysis tools to provide the right information to the right
mation systems literature (e.g. entity-relationship dia- persons throughout the organization with the purpose of
grams, flowcharts and data flow diagrams). Whereas improving strategic and tactical decisions. A BI system is
these modeling techniques can be used when modeling a data-driven DSS that primarily supports the querying of
accounting information systems (Gelinas et al., 2005), But, an historical database and the production of periodic
the REA modeling technique is particular to the AIS do- summary reports (Power, 2008).
main. The REA model, which maps resources, events and Lnnqvist and Pirttimki (2006), stated that the term,
IJSER 2011
292
ISSN 2229-5518
BI, can be used when referring to the following concepts: TABLE 1

1. Related information and knowledge of the or- BI ASSESSMENT CRITERIA
ganization, which describe the business environment, the
organization itself, the conditions of the market, custom-
ers and competitors, and economic issues;
2. Systemic and systematic processes by which or-
ganizations obtain, analyse, and distribute the informa-
tion for making decisions about business operations.
A literature review around the theme of BI shows di-
vision between technical and managerial view points,
tracing two broad patterns. The managerial approach sees
BI as a process in which data, gathered from inside and
outside the enterprise and are integrated in order to gen-
erate information relevant to the decision-making
process. The role of BI here is to create an informational
environment in which operational data gathered from
transactional processing systems (TPS) and external
sources can be analysed, in order to extract strategic
business knowledge to support the unstructured deci-
sions of management.
The technical approach considers BI as a set of tools
that support the process described above. The focus is not
on the process itself, but on the technologies, algorithms
and tools that allow the saving, recovery, manipulation
and analysis of data and information (Petrini and Pozze- ysis allows you to study the properties of questionnaire
bon, 2008). and the items that make them up. The reliability analysis
However, in the overall view, there are two important procedure calculates a number of commonly used meas-
issues. First, the core of BI is the gathering, analysis and ures of scale reliability and also provides information
distribution of information. Second, the objective of BI is about the relationships between individual items in the
to support the strategic decision-making process. measurement scale (Hair et al., 1998).
By strategic decisions, we mean decisions related to The fifth and sixth stages of research framework are
implementation and Assessment of organizational vision, based on factor analysis and are concentrated on extrac-
mission, goals, and objectives, which are supposed to tion and identification of the BI Assessment criteria affect-
have medium- to long-term impact on the organization, ing the intelligence of AIS. Factor analysis is also known
as opposed to operational decisions, which are day-to- as a generic name given to a class of multivariate statistic-
day in nature and more related to execution (Petrini and al methods whose primary purpose is to define the un-
Pozzebon, 2008). derlying structure in a data matrix. With factor analysis,
the researcher can first identify the separate factors of the
structure and then determine the extent to which each
4 RESEARCH METHODOLOGY
variable is explained by each factor. Once these factors
Based on literature review, the points discussed above, and the explanation of each variable are determined, the
the authors recent researches on BI and applying some two primary uses for factor analysis-summarization and
statistical methods, the research structure of this study data reduction-can be achieved. In summarizing the data,
has been developed on seven stages as shown in Figure factor analysis derives underlying factors that, when in-
1. In this way, at first stage a literature review was done terpreted and understood, describe the data in a much
on business intelligence specifications or criteria that a smaller number of concepts than the original individual
system should have to cover BI definitions. These crite- variables (Hair et al., 1998). Evaluating the suitability of
ria listed in Table 1. collected data, performing factor analysis and naming the
At the Second stage, a questionnaire was designed extracted factors are different steps.
with two main parts: first section of the questionnaire
consisted of some questions the characteristics of the in-
terviewees. The content of second section was based on 5 DISCUSSION
business intelligence specifications which were asked as A questionnaire was designed and structured in three
the important evolution criteria. sections. Information related to the basic profile of the
At the third and forth stage, the survey is run to collect interviewees was requested at the beginning of the
data from interviewees; and based on the collected data; questionnaire.
the reliability analysis can be performed. Reliability anal- In the second part, there were 23 questions designed to
IJSER 2011
293
ISSN 2229-5518
measure their attitude, based on the BI Assessment Crite- On the subject of decision-type, the majority of inter-
ria listed in Table 2. The selected response was evaluated viewees make semi-structured and unstructured deci-
by a Likert Scale (Likert, 1974) and the responses could sions in their work. Table 2 also shows the seniority of the
be: very strongly disagree, strongly disagree, disagree, no participants. As can be seen, 20.4 per cent have over 15
opinion, agree, strongly agree or very strongly agree. In years of seniority, 36.4 per cent have 10-15 years, and 43.2
other words, the second part of questionnaire measures per cent have less than 10 years seniority.
their opinions the importance of each BI specification in
terms of the intelligence Assessment Criteria of AIS.
The main targets of the sampling were accounting 5.4 Factor Extaction and Labeling
managers who are involved in systems efforts and deci- Factor analysis can be utilized to examine the underlying
sion-making. patterns or relationships for a large number of variables
and to determine whether the information can be con-
densed or summarized in a smaller set of factors or com-
5.1 Reliability analysis ponents (Hair et al., 1998).
With reliability analysis, you can get an overall index of An important tool in interpreting factors is factor rota-
the repeatability or internal consistency of the measure- tion. The term rotation means exactly what it implies.
ment scale as a whole, and you can identify problem Specifically, the reference axes of the factors are turned
items that should be excluded from the scale. The Cron- about the origin until some other position has been
bachs is a model of internal consistency, based on the reached. The un-rotated factor solutions extract factors in
average inter-item correlation. The Cronbachs (Likert, the order of their importance. The first factor tends to be a
1974) calculated from the 34 variables of this research was general factor with almost every variable loading signifi-
0.941 (94 percent), which showed high reliability for de- cantly, and it accounts for the largest amount of variance.
signed measurement scale. The second and subsequent factors are then based on the
residual amount of variance. The ultimate effect of rotat-
ing the factor matrix is to redistribute the variance from
5.2 Data collection earlier factors to later ones to achieve a simpler, theoreti-
The research targets were accounting managers who were cally more meaningful factor pattern. The simplest case of
involved in systems efforts and decision-making in or- rotation is an orthogonal rotation, in which the axes are
ganizations. The number of questionnaires sent out was maintained at 900 (Hair et al., 1998).
210 and the number returned was 176, which showed a In order to determine whether the partial correlation of
return rate of 83 per cent. the variables is small, the Kaiser-Meyer-Olkin was used to
measure of sampling adequacy (Kaiser, 1958) and Bar-
tletts test of Sphericity (Bartlett, 1950) before starting
5.3 Demographic profiles of interviewees
the factor analysis. The result was a KMO of 0.963 and
The demographic profile of interviewees who participate Bartlett test p-value less than 0.05, which showed good
in the survey has been summarized in Table 2. The results correlation. The factor analysis method is principle com-
show that most of the members (87.5 per cent) are male. ponent analysis in this research, which was developed
Most of the interviewees (88.7 per cent) have a Bachelor of by Hotteling (1935). The condition for selecting factors
Science (BS) or a higher degree, as shown in Table 2. was based on the principle proposed by Kaiser (1958):
TABLE 2 TABLE 3
DEMOGRAPHIC PROFILES OF INTERVIEWEES ROTATED FACTOR ANALYSIS RESULTS
Eigen value larger than one, and an absolute value of fac-

tor loading greater than 0.5. The 23 variables were
grouped into five factors. The results can be seen in Table
3. Five factors had an Eigen value greater than one and
the interpretation variable was 69.349 percent. The factors
IJSER 2011
294
ISSN 2229-5518
were rotated according to Varimax rotation method. [5] Azadivar, F., Truong, T. and Jiao, Y., 2009. A decision support
To indicate the meaning of the factors, they have been system for fisheries management using operations research and
given short labels indicating their content. Analytical systems science approach, Expert Systems with Applications
Decision-support, Providing Integration with Envi- 36: 29712978.
ronmental and Experimental Information, Optimiza- [6] Bartlett, M.S., 1950. Test of significance in factor analysis, Brit-
tion Model, Reasoning and finally, Enhanced Deci- ish Journal of Psychology, Vol. 3, 77-85.
sion-making Tools are the names which have been as- [7] Berson, A., & Smith, S., 1997. Data warehousing, data mining,
signed to the extracted factors. and OLAP: McGraw-Hill, Inc. New York, NY, USA.
[8] Birnberg, J. G., & Shields, J. F. (1989). Three decades of beha-
vioral accounting research: A search for order. Beha-vioral Re-
search in Accounting, 1, 2374.
6 CONCLUSION [9] Bolloju, N., Khalifa, M. and Turban, E., 2002. Integrating know-
Enterprise systems like Accounting Information Systems ledge management into enterprise environments for the next
(AIS) converts and store the data. Therefore, it is impor- generation decision support, Decision Support Systems 33: 163
tant to integrate decision-support into the environment of 176.
these systems. Business intelligence (BI) can be embedded [10] Bose, R., 2009. Advanced analytics: opportunities and chal-
in these enterprise systems to obtain competitive advan- lenges, Industrial Management & Data Systems, Vol. 109, No. 2,
tage. 155-172.
This research confirmed the necessity to assess the In- [11] Bouwens, J., & Abernethy, M. A. (2000). The Consequences of
telligence of Accounting Information Systems and dem- Customization on Management Accounting System Design.
onstrated that this assessment can advance a decision- Accounting, Organizations and Society, 25(3), 221241.
support environment. From a wide-ranging literature [12] Courtney, J.F., 2001. Decision making and knowledge man-
review, 23 criteria for BI assessment were gathered and agement in inquiring organizations: toward a new decision-
embedded in the second part of the research. The inter- making paradigm for DSS, Decision Support Systems 31: 1738.
viewees selected the more important criteria from these [13] Damart, S., Dias, L. and Mousseau, V., 2007. Supporting groups
23 variables by assigning ranks to them. The research in sorting decisions: Methodology and use of a multi-criteria
then applied factor analysis to extract the five factors for aggregation/disaggregation DSS, Decision Support Systems 43:
evaluation. These factors were Analytical Decision- 14641475.
support, Providing Integration with Environmental and [14] David JS, Gerard GJ,McCarthyWE. Design science: anREA-
Experimental Information, Optimization Model, Rea- perspective on the future of AIS. In:ArnoldV, Sutton SG, edi-
soning and finally, Enhanced Decision-making Tools. tors. Researching accounting as an information systems discip-
The authors believe that after this research, organizations line. Sarasota, FL, USA: American AccountingAssociation; 2002.
can make better decisions for designing, selecting, evalu- [15] Delorme, X., Gandibleux, X. and Rodrguez, J., 2009. Stability
ating and buying Accounting Information Systems, using evaluation of a railway timetable at station level, European
criteria that help them to create a better decision-support Journal of Operational Research 195: 780790.
environment in their work systems. Of course, further [16] Denna EL, Cherrington JO, Andros DP, Hollander AS. Event-
research is needed. One area is the design of expert sys- driven business solutions. Homewood, IL, USA: Business One
tems (tools) to compare vendors products. The other is Irwin; 1993.
the application these criteria and factors in a Multy Crite- [17] Eckerson , W., 2005. Performance dashboards: Measuring, mon-
ria Decision Making framework to select and rank AIS, itoring, and managing your business. Wiley Press.
financial and banking systems based on BI specification. [18] Elbashir, M., Collier, P. and Davern, M., 2008. Measuring the
The complex relationship between decision-making satis- effects of business intelligence systems: The relationship be-
faction of managers, and these factors should also be ad- tween business process and organizational performance. Inter-
dressed in future research. national Journal of Accounting Information Systems, 9(3), 135-
153.
[19] Eom, S., 1999. Decision support systems research: current state
REFERENCES and trends, Industrial Management & Data Systems 99/5
213220.
[1] Abernethy MA, Vagnoni E. Power, 2004. Organization design
[20] Evers, M., 2008. An analysis of the requirements for DSS on
and managerial behaviour. Account Organ Soc ;29(3/4):20725.
integrated river basin management, Management of Environ-
[2] Alter, S., 2004. A work system view of DSS in its fourth decade.
mental Quality: An International Journal Vol. 19 No. 1, pp. 37-
Decision Support Systems, 38(3), 319-327.
53.
[3] Anders Rom, Carsten Rohde, 2007. Management accounting
[21] Fazlollahi, B. and Vahidov, R., 2001. Extending the effectiveness
and integrated information systems: A literature review, Inter-
of simulation-based DSS through genetic algorithms, Informa-
national Journal of Accounting Information Systems 8 4068
tion & Management 39: 53-65.
[4] Arnold V, Collier PA, Leech SA, Sutton SG. Impact of intelli-
[22] Feng, Y., Teng, T. and Tan, A., 2009. Modelling situation
gent decision aids on expert and novice decision-
awareness for Context-aware Decision Support, Expert Systems
makers'judgments. Account Finance 2004;44(1):126.
IJSER 2011
295
ISSN 2229-5518
with Applications 36: 455463. Mitra, G., 2000. A prototype decision support system for stra-
[23] Galasso, F. and Thierry, C., 2008. Design of cooperative tegic planning under uncertainty, International Journal of Phys-
processes in a customersupplier relationship: An approach ical Distribution & Logistics Management, Vol. 30 No. 7/8, 640-
based on simulation and decision theory, Engineering Applica- 660.
tions of Artificial Intelligence. [42] Kwon, O., Kim, K. and Lee, K.C., 2007. MM-DSS: Integrating
[24] Gao, S. and Xu, D., 2009. Conceptual modeling and develop- multimedia and decision-making knowledge in decision sup-
ment of an intelligent agent-assisted decision support system port systems, Expert Systems with Applications 32: 441457.
for anti-money laundering, Expert Systems with Applications [43] Lamptey, G., Labi, S. and Li, Z., 2008. Decision support for
36: 14931504. optimal scheduling of highway pavement preventive mainten-
[25] Geerts GL, McCarthy WE. An ontological analysis of the eco- ance within resurfacing cycle, Decision Support Systems 46:
nomic primitives of the extended REA enterprise information 376387.
architecture. Int J Account Inf Syst 2002;3(1):116. [44] Lee, J. and Park, S., 2005. Intelligent profitable customers seg-
[26] Gelinas UJ, Sutton SG, Hunton JE. Accounting information mentation system based on business intelligence tools, Expert
systems. 6th edition. Thomson, OH, USA: South-Western; Systems with Applications 29: 145152.
[27] 2005. [45] Li, D., Lin, Y. and Huang, Y., 2009. Constructing marketing
[28] Ghoshal, S. and Kim, S.K., 1986. Building Effective Intelligence decision support systems using data diffusion technology: A
Systems for Competitive Advantage, Sloan Management Re- case study of gas station diversification, Expert Systems with
view, Vol. 28, No. 1, 4958. Applications 36: 25252533.
[29] Gonnet, S., Henning, G. and Leone, H., 2007. A model for cap- [46] Li, S., Shue, L. and Lee, S., 2008. Business intelligence approach
turing and representing the engineering design process, Expert to supporting strategy-making of ISP service management, Ex-
Systems with Applications 33: 881902. pert Systems with Applications 35: 739754.
[30] Gonzlez, J.R., Pelta, D.A. and Masegosa, A.D., 2008. A frame- [47] Likert, R., 1974. The method of constructing an attitude scale, in
work for developing optimization-based decision support sys- Maranell, G.M. (Ed.), Scaling: A Sourcebook for Behavioral
tems, Expert Systems with Applications. Scientists, Aldine Publishing Company, Chicago, IL, 21-43.
[31] Gottschalk, P., 2006. Expert systems at stage IV of the know- [48] Lin, Y., Tsai, K., Shiang, W., Kuo, T. and Tsai, C., 2009. Research
ledge management technology stage model: The case of police on using ANP to establish a performance assessment model for
investigations, Expert Systems with Applications 31: 617628. business intelligence systems, Expert Systems with Applica-
[32] Goul, M. and Corral, K., 2007. Enterprise model management tions 36: 41354146.
and next generation decision support, Decision Support Sys- [49] Loebbecke, C. and Huyskens, C., 2007. Development of a mod-
tems 43: 915 932. el-based netsourcing decision support system using a five-stage
[33] Gngr Sen, C., Baral, H., Sen, S. and Baslgil, H., 2008. An methodology, European Journal of Operational Research.
integrated decision support system dealing with qualitative [50] Lnnqvist, A. and Pirttimki, V., 2006. The Measurement of
and quantitative objectives for enterprise software selection, Business Intelligence, Information Systems Management, 23:1,
Expert Systems with Applications. 3240.
[34] Hair, J.F., Anderson, R.E., Tatham, R.L. and Black, W.C., 1998. [51] Makropoulos, C.K., Natsis, K., Liu, S., Mittas, K. and Butler, D.,
Multivariate Data Analysis, Prentice-Hall, Upper Saddle River, 2008, Decision support for sustainable option selection in inte-
NJ, 7-232. grated urban water management, Environmental Modelling &
[35] Hedgebeth, D., 2007. Data-driven decision making for the en- Software 23: 14481460.
terprise: an overview of business intelligence applications, The [52] Maria, F., 2005. Improving the utilization of external strategic
journal of information and knowledge management systems information. Tampere University of Technology, Master of
Vol. 37 No. 4, 414-420. Science Thesis.
[36] Hemsley-Brown, J., 2005. Using research to support manage- [53] Marinoni, O., Higgins, A., Hajkowicz, S. and Collins, K, 2009.
ment decision making within the field of education, Manage- The multiple criteria analysis tool (MCAT): A new software tool
ment Decision Vol. 43 No. 5, pp. 691-705. to support environmental investment decision making, Envi-
[37] Hewett, C., Quinn, P., Heathwaite, A.L., Doyle, A., Burke, S., ronmental Modelling & Software 24: 153164.
Whitehead, P. and Lerner, D., 2009. A multi-scale framework [54] Metaxiotis, K., Psarras, J. and Samouilidis, E., 2003, Integrating
for strategic management of diffuse pollution, Environmental fuzzy logic into decision support systems: current research and
Modelling & Software 24: 7485. future prospects, Information Management & Computer Securi-
[38] Hotteling, H., 1935. The most predictable criterion, Journal of ty 11/2, 53-59.
Educational Psyhology, Vol. 26, 139-142. [55] McCarthy WE. An entity-relationship view of accounting mod-
[39] Kaiser, H., 1958. The varimax criterion for analytic rotation in els. Account Rev 1979;54(4):66786.
factor analysis. Psychometrika, 23(3), 187-200. [56] McCarthy WE. The REA accounting model: a generalized
[40] Koo, L.Y., Adhitya, A., Srinivasan, R. and Karimi, I.A., 2008. framework for accounting systems in a shared data environ-
Decision support for integrated refinery supply chains Part 2. ment. Account Rev 1982;57(3):55478.
Design and operation, Computers and Chemical Engineering [57] Nemati, H., Steiger, D., Iyer, L. and Herschel, R., 2002. Know-
32: 27872800. ledge warehouse: an architectural integration of knowledge
[41] Koutsoukis, N., Dominguez-Ballesteros, B., Lucas, C.A. and management, decision support, artif icial intelligence and data
IJSER 2011
296
ISSN 2229-5518
warehousing, Decision Support Systems 33: 143 161. decision support system for managing inventory at GlaxoS-
[58] Nie, G., Zhang, L., Liu, Y., Zheng, X. and Shi, Y., 2008. Decision mithKline, Decision Support Systems.
analysis of data mining project based on Bayesian risk, Expert [76] Shi, Z., Huang, Y., He, Q., Xu. L., Liu, S., Qin, L., Jia, Z., Li, J.,
Systems with Applications. Huang, H. and Zhao, L., 2007.MSMinera developing platform
[59] Noori, B. and Salimi, M.H., 2005. A decision-support system for for OLAP, Decision Support Systems 42: 2016 2028.
business-to-business marketing, Journal of Business & Industri- [77] Shim, J., Warkentin,M., Courtney, J., Power, D., Sharda, R. and
al Marketing 20/4/5: 226236. Carlsson, C., 2002. Past, present, and future of decision support
[60] Nylund, A., 1999. Tracing the BI Family Tree. Knowledge Man- technology, Decision Support Systems 33: 111 126.
agement. [78] Sutton SG, Arnold V. Foundations and frameworks for AIS
[61] O zbayrak, M. and Bell, R., 2003. A knowledge-based decision research. In: Arnold V, Sutton SG, editors. Researching account-
support system for the management of parts and tools in FMS, ing as an information systems discipline. Sarasota, FL, USA:
Decision Support Systems 35: 487 515. American Accounting Association; 2002.
[62] Petrini, M. and Pozzebon, M., 2008. What role is Business [79] Wadhwa, S., Madaan, J. and Chan, F.T.S., 2009. Flexible deci-
Intelligence playing in developing countries? A picture of Bra- sion modeling of reverse logistics system: A value adding
zilian companies. In: Rahman, Hakikur (Eds.), Data Mining MCDM approach for alternative selection, Robotics and Com-
Applications for Empowering Knowledge Societies, IGI Global, puter-Integrated Manufacturing 25: 460469.
237257. [80] Yu, L., Wang, S. and Lai, K., 2009. An intelligent-agent-based
[63] Phillips-Wren, G., Hahn, E. and Forgionne, G., 2004. A mul- fuzzy group decision making model for financial multicriteria
tiple-criteria framework for evaluation of decision support sys- decision support: The case of credit scoring, European Journal
tems, Omega 32: 323 332. of Operational Research 195: 942959.
[64] Phillips-Wren, G., Mora, M., Forgionne, G.A. and Gupta, J.N.D., [81] Zack, M., 2007. The role of decision support systems in an inde-
2007. An integrative evaluation framework for intelligent deci- terminate world, Decision Support Systems 43: 16641674.
sion support systems, European Journal of Operational Re- [82] Zhan, J., Loh, H.T. and Liu, Y., 2009. Gather customer concerns
search. from online product reviews A text summarization approach,
[65] Pitty, S., Li, W., Adhitya, A., Srinivasan, R. and Karimi, I.A., Expert Systems with Applications 36: 21072115.
2008. Decision support for integrated refinery supply chains [83] Zhang, X., Fu, Z., Cai, W., Tian, D. and Zhang, J.,
Part 1. Dynamic simulation, Computers and Chemical Engi- 2009. Applying evolutionary prototyping model in
neering 32: 27672786. developing FIDSS: An intelligent decision support
[66] Plessis, T. and Toit, A.S.A., 2006. Knowledge management and system for fish disease/health management, Expert
legal practice, International Journal of Information Manage- Systems with Applications 36: 39013913.
ment 26: 360371.
[67] Power, D. and Sharda, R., 2007. Model-driven decision support
systems: Concepts and research directions, Decision Support
Systems 43: 1044 1061.
[68] Power, D.J., 2008.Understanding Data-Driven Decision Support
Systems, Information Systems Management, 25:2, 149 154.
[69] Quinn, N.W.T., 2009. Environmental decision support system
development for seasonal wetland salt management in a river
basin subjected to water quality regulation, agricultural water
management 96, 247 254.
[70] Raggad, B.G., 1997. Decision support system: use IT or skip IT,
Industrial Management & Data Systems 97/2: 4350.
[71] Ranjan, J., 2008. Business justification with business intelli-
gence, VINE: The journal of information and knowledge man-
agement systems, Vol. 38, No. 4, 461-475.
[72] Reich, Y. and Kapeliuk, A., 2005. A framework for organizing
the space of decision problems with application to solving sub-
jective, context-dependent problems, Decision Support Systems
41: 1 19.
[73] Ross, J.J., Dena, M.A. and Mahfouf, M., 2008. A hybrid hierar-
chical decision support system for cardiac surgical intensive
care patients. Part II. Clinical implementation and evaluation,
Artificial Intelligence in Medicine.
[74] Santhanam, R. and Guimaraes, T., 1995. Assessing the quality
of institutional DSS, European Journal of Information Systems 4
(3).
[75] Shang, J., Tadikamalla, P., Kirsch, L. and Brown, L., 2008. A
IJSER 2011
297
ISSN 2229-5518
Processing of Images Based on

Segmentation Models for Extracting
Textured Component
V M Viswanatha, Nagaraj B Patil, Sanjay Pande MB
Abstract The method for segmentation of color regions in images with textures in adjacent regions being different can be arranged in two
steps namely color quantization and segmentation spatially. First, colors in the image are quantized to few representative classes that can be
used to differentiate regions in the image. The image pixels are then replaced by labels assigned to each class of colors. This will form a
class-map of the image. A mathematical criteria of aggregation and mean value is calculated. Applying the criterion to selected sized windows
in the class-map results in the highlighted boundaries. Here high and low values correspond to possible boundaries and interiors of color
texture regions. A region growing method is then used to segment the image.
Key Words: Texture segmentation, clustering, spital segmentation, slicing, texture composition, boundry value image, median-cut.
1. INTRODUCTION
S egmentation is the low-level operation concerned with

partitioning images by determining disjoint and ho-
processing color images requires computation times con-
siderably larger than those needed for gray-level images.
mogeneous regions or, equivalently, by finding edges or With an increasing speed and decreasing costs of compu-
boundaries. Regions of an image segmentation should be tation; relatively inexpensive color camera the limitations
uniform and homogeneous with respect to some charac- are ruled out. Accordingly, there has been a remarkable
teristics such as gray tone or texture Region interiors growth of algorithms for segmentation of color images.
should be simple and without many small holes. Adjacent Most of times, these are kind of dimensional exten-
regions of segmentation should have significantly differ- sions of techniques devised for gray-level images; thus
ent values with respect to the characteristic on which they exploit the well-established background laid down in that
are uniform. Boundaries of each segment should be sim- field. In other cases, they are ad hoc techniques tailored
ple, not ragged, and must be spatially accurate. on the particular nature of color information and on the
Thus, in a large number of applications in image physics of the interaction of light with colored materials.
processing and computer vision, segmentation plays a More recently, Yining Deng and B. S. Manjunath [1][2]
fundamental role as the first step before applying the uses the basic idea of separate the segmentation process
higher-level operations such as recognition, semantic in- into color quantization and spatial segmentation. The
terpretation, and representation. quantization is performed in the color space without con-
Earlier segmentation techniques were proposed sidering the spatial distributions of the colors. S Belongie,
mainly for gray-level images on which rather comprehen- et.al [3], in their paper present a new image representa-
sive survey can be found. The reason is that, although tion, which provides a transformation from the raw pixel
color information permits a more complete representation data to a small set of image regions that are coherent in
of images and a more reliable segmentation of them, color and texture space. A new method of color image
IJSER 2011
298
ISSN 2229-5518
segmentation is proposed in [4] based on K-means algo- assured of using every color in the colormap, and thereby
rithm. Both the hue and the intensity components are ful- reproducing the original image more closely.
ly utilized.
The most popular algorithm for color quantization,
invented by Paul Heckbert in 1980, is the median cut
algorithm. Many variations on this scheme are in use.

2 COLOR QUANTIZATION
Before this time, most color quantization was done using
Color Quantization is a form of image compres- the popularity algorithm, which essentially constructs a
sion that reduces the number of colors used in an image histogram of equal-sized ranges and assigns colors to the
while maintaining, as much as possible, the appearance of ranges containing the most points. A more modern
the original. The optimal goal in the color quantization popular method is clustering using Octree,
process is to produce an image that cannot be distin-
guished from the original. This level of quantization in 3. Segmentation
fact may never be achieved. Thus, a color quantization The division of an image into meaningful struc-
algorithm attempts to approximate the optimal solution. tures, image segmentation, is often an essential step in im-
The process of color quantization is often bro- age analysis. A great variety of segmentation methods
ken into four phases . have been proposed in the past decades. They can be ca-
1) Sample image to determine color distribution. tegorized into
2) Select colormap based on the distribution Threshold based segmentation: Histogram thresholding
3) Compute quantization mapping from 24-bit colors to and slicing techniques are used to segment the image.
representative colors Edge based segmentation: Here, detected edges in an
4) Redraw the image, quantizing each pixel. image are assumed to represent object boundaries, and
Choosing the colormap is the most challenging used to identify these objects.
task. Once this is done, computing the mapping table Region based segmentation: Here the process starts in
from colors to pixel values is straightforward. the middle of an object and then grows outwards until it
In general, algorithms for color quantization can meets the object boundaries.
be broken into two categories: Uniform and Non- Clustering techniques: Clustering methods attempt to
Uniform. In Uniform quantization the color space is bro- group together patterns that are similar in some sense.
ken into equal sized regions where the number of regions Perfect image segmentation cannot usually be
NR is less than or equal to Colors K. Uniform quantiza- achieved because of oversegmentation or undersegmentation.
tion, though computationally much faster, leaves much In oversegmentation pixels belonging to the same object
room for improvement. In Non-Uniform quantization the are classified as belonging to different segments.
manner in which the color space is divided is dependent In the latter case, pixels belonging to different
on the distribution of colors in the image. By adapting a objects are classified as belonging to the same object.
colormap to the color gamut of the original image, it is
IJSER 2011
299
ISSN 2229-5518
4. Design local homogeneities at different scales, which can indicate

Natural pictures are rich in color and texture. potential boundary locations.
Texture segmentation algorithms require the estimation A spatial segmentation algorithm that grows regions
of texture model parameters which is difficult. The goal of from seed areas of the Boundary-images to achieve the
this work is to segment images into homogeneous color- final segmentation. Figure. 4.1 shows a schematic of the
texture regions. The approach tests for the homogeneity algorithm for color image segmentation.
of a given color-texture pattern, which is computationally
Color Image
more feasible than model parameter estimation. In order
to identify homogeneity, the following assumptions are
made: COLOR SPACE QUANTIZATION
Each image contains homogeneous color-texture re-

gions. Class-map of Color Image
The information in each image region can be

represented by a set of few quantized colors.
Spatial Segmentation
The colors between two neighboring regions are dis-
tinguishable
Boundary calculation of Image
4.1 The Process of Segmentation

Boundary-image
The segmentation is carried out in two stages. In
the first stage, colors in the image are quantized to several
representative classes to differentiate regions in the im-
Region Growing
age. Then the image pixel values are replaced by their
corresponding labels to form a image class-map that can

Segmentation Results
be viewed as a special kind of texture composition. In the
second stage, spatial segmentation is performed directly Fig.4.1: Schematic of Segmentation

class-map. A few good existing quantization algorithms
are used in this work.

4.2 The Parameter Calculation as a criterion for
The focus of this work is on spatial segmentation Segmentation
and can be summarized as: It is difficult to handle a 24-bit color images with
For image segmentation a parameter is calculated. This thousands of colors. Image are coarsely quantized with-
involves minimizing a cost associated with the partition- out significantly degrading the color quality. Then, the
ing of the image based on pixel labels. quantized colors are assigned labels. A color class is the
Segmentation is achieved using an algorithm. The nota- set of image pixels quantized to the same color. The image
tion of Boundary-images, correspond to measurements of pixel colors are replaced by their corresponding color
class labels and this image is called a class-map. Usually,

IJSER 2011
300
ISSN 2229-5518
each image region contains pixels from a small subset of two regions. One region contains class 1 and the other
the color classes and each class is distributed in a few im- one contains classes 2 and 0.
age regions. Now J is recalculated over each segmented re-

The class-map image is viewed as a special kind gion instead of the entire class-map and define the aver-
of texture composition. The value of each point in the age Javg by
class-map is the image pixel position, a 2-D vector (x, y).

Javg = 1/N MkJk
Each point belongs to a color class.
Let Z be the set of all N data points in a class- where Jk represents J calculated over region k, Mk is the
map. Let , z= (x,y), z Z , and m be the mean, number of points in region k, N is the total number of
m = 1/N z points in the class-map, and the summation is over all the
zZ regions in the class-map. Javg is the parameter used for

Suppose Z is classified into C classes, Zi, i =1,C. Let mi segmentation. For a fixed number of regions, a better
be the mean of the Ni data points of class Zi, segmentation tends to have a lower value of Javg . If the
mi = 1/Ni z segmentation is good, each segmented region contains a

z Zi few uniformly distributed color class labels and the re-
Let sulting J value for that region is small. Therefore, the

overall Javg is also small. Notice that the minimum value
ST = || z-m || 2
zZ Javg can take is 0. For the case of class-map 1, the partition-
ing shown is optimal(Javg =0).
SW = Si = || z-mi || 2 Figure.4.2 shows the manual segmentation results of

i = 1 i =1 z Zi class-maps.
Since the global minimization of Javg for the entire
SW is the total variance of points belonging to the same image is not practical , J, if applied to a local area of the
class. class-map, indicates whether that area is in the region
interiors or near region boundaries
J = ST SWSW
J=1.72 J=0.855
For the case of an image consisting of several homo- J1 = 0, J2 = 0, J0 = 0 J1 = 0, J{2,0o} = 0.011

geneous color regions, the color classes are more sepa- Javg = 0 Javg = 0.05
rated from each other and the value of J is large. An ex-

11111 0000 11111 2020
ample of such a class-map is class-map 1 in Figure.4.2 and 11111 0000 11111 0202
11111 0000 11111 2020
the corresponding J value is 1.7. Figure. 4.2 shows anoth-
11111 0000 11111 0202
er example of a three class distribution for which J is 0.8, 11111 0000 11111 2020
1111 22222 1111 20202
indicating a more homogeneous distribution than class- 1111 22222 1111 02020
1111 22222 1111 20202
map 1. For class-map 2, a good segmentation would be
1111 22222 1111 02020
IJSER 2011
301
ISSN 2229-5518
tized into along with the algorithm to choose from be-

Figure 4.2: Segmented class-maps and their corresponding tween Median-Cut, NeuQuant, Oct-Tree.
J values. The module Segmentor gets arguments regarding the

Now an image is constructed whose pixel values initial window size and maximum number of iterations.
correspond to these J values calculated over small win- Here the segmentation parameter is calculated and boun-
dows centered at the pixels. These are refered as Boun- dary image is created. It scans the color vector in boun-
dary-values Images and the corresponding pixel values as dary image and calculates initial segmented image. Then
local J values. The higher the local J value is, the more for the number of iterations mentioned it segments im-
likely that the corresponding pixel is near a region boun- ages repeatedly by converging the values. The accuracy of
dary. The Boundary-image contains intensities that actual- segmentation depends on the number of iterations the
ly represent the region interiors and region boundaries, image is segmented.

respectively. Windows of small size are useful in localiz-
ing the intensity/color edges, while large windows are The module for Region Growing implements a
useful for detecting texture boundaries. Often, multiple simple region growing algorithm. It runs a classic stack-
scales are needed to segment an image. In this implemen- based region rowing algorithm. It finds a pixel that is not
tation, the basic window at the smallest scale is a 9 x 9 labeled, labels it and stores its coordinates on a stack.
window. The smallest scale is denoted as scale 1. The While there are pixels on the stack, it gets a pixel from the
window size is doubled each time to obtain the next larg- stack (the pixel being considered), Checks its neighboring
er scale. pixels to see if they are unlabeled and close to the consi-
The characteristics of the Boundary-images allow dered pixel, if are, label them and store them on the stack
us to use a region-growing method to segment the image. Repeats this process until there are no more pixels on the
Initially the entire image is considered as one region. The image.
segmentation of the image starts at a coarse initial scale. It Figure 5.1: Snapshot of Selected Image and Segmented Image
(NeuQuant,12)
then repeats the same process on the newly segmented
regions at the next finer scale. Region growing consists of
determining the seed points and growing from those seed
locations. Region growing is followed by a region merg-

ing operation to give the final segmented image.
5. Implementation
The implementation of segmentation is carried
out using JDK 1.5. and JAI 1.3 (Java Advanced Imaging
1.3) on windows platform.

The module ColorQuantizer implements a GUI
that takes the number of colors the image should be quan-
IJSER 2011
302
ISSN 2229-5518
also changes the result. The number of colors for quanti-

zation chosen and the number of iterations changes the
number of regions in the segmented image.

6. Conclusion
The color image is quantized and then is seg-
mented with a focus on spatial segmentation. The thre-

shold values are calculated that is used for the segmenta-
tion at various scales. At each step the segments are iden-

tified through the boundaries and valleys that are identi-
fied with the threshold values that are low near the
boundary.The accuracy of segmentation depends on the
algorithm used for quantization and the number of itera-
tions.
Figure 5.3: Snapshot of Selected Image and Segmented
Figure 5.2: Snapshot of Selected Image and Segmented Image (Mediun Cut,8)
Image (NeuQuant,16)
Once the segmentation is over the image is displayed
In Figures 5.1 thru 5.3 ,we can see the border dis- with appropriate borders to differentiate between the
played on the original image, border displayed on the segments. The images having different colors in the adja-
segmented image, original image and segmented image , cent regions are the best candidates for the segmentation
shown in top left to bottom right in that order using the proposed method.
It is observed that changing of the parameters

IJSER 2011
303
ISSN 2229-5518
However, it has does not handle pictures where [4] Pappas, T.N., An adaptive clustering algorithm for
smooth transition takes place between adjacent regions image segmentation, International Conference on
and there is no clear visual boundary. For instance, the Pattern recognition, Vol 3, pp 3619, 2000.
color of a sunset sky can vary from red to orange to dark [5] H.D. Cheng, X,H Jiang, Y Sun and Jingli Wang,
in a very smooth transition. Color Image Segmentation: Advances and Pros-
pects, IEEE Transactions on Pattern Recognition, Vol
7. Future Work 34, pp 2259- 2281, December 2001.

[6] D.K. Panjwani and G. Healey, Markov random field
The algorithm can be modified to find out the number
models for unsupervised segmentation of textured
of colors adaptively for quantization instead of taking it
color images, IEEE Trans. on Pattern Analysis and
as a parameter.
Machine Intelligence, vol. 17, no.10, p. 939-54, 1995.
The algorithm can be extended to handle images with [7]
Heng-Da Cheng Ying Sun , A hierarchical ap-
smooth transitions between adjacent regions. The method proach to color image segmentation using homogene-
can be extended to segment the color-texture regions in
ity, Vol 9, pp 2071- 2082.
video data.
[8] Jianbo Shi, Jitendra Malik,Normalized cuts and
ACKNOWLEDGEMENTS image segmentation, IEEE Transactions on Pattern
The authors like to thank Singhania Universaity, Rajas- Analysis and Machine Intelligence ,Vol 22, pp 888-
than and VVIT Institute of Technology, Mysore. 905, August 2000.

[9] O'Reilly - Safari Books Online - Java 2D Graphics.
BIBLIOGRAPHY
[10] http://java.sun.com/products/java-
[1] Y. Deng, B. S. Manjunath, "Unsupervised segmenta- media/jai/forDevelopers/
tion of color-texture regions in images and video," [11] Digital Image Processing, Rafael C. Gonzalez, Ri-
IEEE Transactions on Pattern Analysis and Machine chard E.Woods, Pearson Education, Addison Wes-
Intelligence , Aug 2001. ley Longman
[2] Y. Ma and B. S. Manjunath, "EdgeFlow: A technique [12] Digital Image Processing, a practical introduction
for boundary detection and segmentation", IEEE using Java, Nick Efford, Pearson Education
Transactions on Vol. 9, pp. 1375-88, August 2000.
[3] S.Belongie, C. Carson, H.Greenspan and
J.Malik,Color and texture based image segmentation

usingEM and its application to content based image
retrieval, International Conference on Computer Vi-
sion, pp. 675-682, 1998.
IJSER 2011
304

ISSN 2229-5518
An Adaptive and Efficient XML Parser Tool for

Domain Specific Languages
W. Jai Singh, S. Nithya Bala
Abstract XML (eXtensible Markup Language) is a standard and universal language for representing information. XML has become integral to
many critical enterprise technologies with its ability to enable data interoperability between applications on different platforms. Every application that
processes information from XML documents needs an XML Parser which reads an XML document and provides interface for user to access its con-
tent and structure. However the processing of xml documents has a reputation of poor performance and a number of optimizations have been de-
veloped to address this performance problem from different perspectives, none of which have been entirely satisfactory. Hence, in this paper we
developed a Fast Parser tool for domain specific languages. In this we can execute Parser user friendly without having any constraints.
Index TermsXML, Parser Tool, Document Object Model, SAX, XML Document, Document Validation.
1 INTRODUCTION
T he XML (eXtensible Markup Language) is now widely

adopted within (networked) applications. Due to its flex-
ibility and efficiency in transmission of data, XML has be-
Several efforts have been made to address the parsing
and validation performance through the use of grammar based
parser generation by leveraging XML schema language such as
come the emerging standard of data transfer and data ex- DTD (Document Type Definition), XML schema at compile
change across the application and Internet [1]. XML has poten- time. DOM (Document Object Model) and SAX (Simple API for
tial as a back end solution as well as a marvelous standard for XML) are the two most widely used XML parsing models, none
re-designing databases and other content. XML has become of which has been entirely satisfactory [2], [5].
integral to many critical enterprise technologies with its ability A parser can read the XML document components via
to enable data interoperability between on different platforms. Application Programming Interfaces (APIs) in two approaches.
With various conversion tools now emerging in the market- For stream-based approach such as SAX (also known as event-
place, XML can be used to bridge between different applica- based parser) and tree-based approach such as DOM. DOM
tions. Its standard-based, set forth a design for structuring fu- (Document Object Model) and SAX (Simple API for XML) are
ture content [1], [2]. the two most widely used XML parsing models, none of which
XML delivers key advantages in interoperability due to its has been entirely satisfactory [2], [3], [5], [7]. A brief description
flexibility, expressiveness and platform-neutrality. As XML has of them is given as follows.
become a performance-critical aspect of the next generation of DOM is a tree-based interface that models an XML
business computing infrastructure. Tomorrows computers will document as a tree of various nodes. The main advantage of
have more cores rather than exponentially faster clock speeds, this parse method is that it supports random access to the doc-
and software will increasingly have to rely on parallelism to ument. DOM parsers create a node object for each node that
take advantage of this trend. precisely models all the structure and content information [2],
Every application that processes information from XML [3], [5]. DOM is an easy way to work with XML. However,
documents needs an XML Parser which reads an XML docu- DOM parsers take too much time and memory, making them
ment and provides interface for user to access its content and unavailable for large XML documents. Moreover, no actual
structure. An XML parser facilitates in simplifying the process work can be done before completely parsing XML, which in-
of manipulating XML documents. There are mainly two chal- troduces significant delay and is unacceptable in enterprise
lenges for generic XML parsers. One is that the code size of the applications.
XML Parsers is restricted because of the limitation of memory. SAX is an event-based parsing model that reads an
The other is the run time adaptability of XML Parsers is re- XML document from beginning to end. Each time it encounters
quired due to the diversity of applications in terms of their de- a syntax construction; it generates an event and notifies the
pendency on XML syntax set. application [2], [5], [7]. It does not preserve the structure and
content information in memory, thus saving a large amount of

memory space. Unfortunately, they lack the ability of random
W. Jai Singh, Assistant Professor, Department of MCA, access and are forward only, which limits their use to a very
Park College of Engineering and Technology, Coimbatore,
Tamil Nadu, India 641 659. small scope. SAX is memory efficient but writing a SAX parser
S. Nithya Bala, PG Scholar, Department of MCA, Park is complex.
College of Engineering and Technology, Coimbatore, Ta- Several methods were presented to improve XML
mil Nadu, India 641 659 parsing from different viewpoints. There are a number of ap-
proaches trying to address the performance bottleneck of XML
parsing. The typical software solutions include the pull-based
IJSER 2011
305

ISSN 2229-5518
parsing [9], lazy parsing [10] and schema-specific parsing [4]. ment. The Fast XML Parser Tool is as follows:
Su Cheng Haw and G. S. V. Radha Krishna Rao have
presented a model called Comparative Study and Benchmark- Load an XML document: Before an XML document can be ac-
ing on XML Parser. In that, they compare the xerces and .NET cessed and manipulated, it must be loaded into an XML Parser.
parsers based on the performance, memory, usage and so on The XML Parser reads XML document and converts it into a
[2]. Giuseppe Psaila have been developed a system called meaningful format. The job of the XML Parser is to make sure
Loosely coupling Java algorithms and XML Parsers. In that, that the document meets the defined structure and constraints.
he conducted a study about the problem of coupling java algo- The validation rule for any particular sequence of character
rithms with XML parsers. Su-cheng Haw and Chien-Sing Lee data is determined by the type of definition for its enclosing
have been presented a model called Fast Native XML Storage element. Fig.2 shows the sample XML document.
and Qurey Retrieval. In that, they proposed the INLAB2 ar-
chitecture comprises of five main components namely XML Algorithm
parser, XML encoder, XML indexer, Data manager and Query Step 1: Read XML File
processor [10]. Fadi El-Hassan and Dan Ionescu presented An Step 2: Search for start tag using regular expression <>
efficient Hardware based XML parsing techniques. In that, Step 3: If a start tag found then
they proposed hardware based solutions can be an obvious Step 4: Add attributes to hash table
choice to parse XML in a very efficient manner. Step 5: Search or end tag using regular expression </>
The existing XML Parsers spend a large amount of Step 6: Add element to hash table
time in tokenizing the input. To overcome all the drawbacks, Step 7: End if
here we have developed a new Fast Parser tool for domain spe- Step 8: Repeat step 3 to 7 until End Of File
cific languages. Though careful analysis of the operations re- Step 9: Verify and validate each element in hash table
quired for parsing and validation, we are using hash table to Step 10: End XML Parser
store element information, this will enhance the speed of acces-
sibility while searching for an element. More over we are using Reading an XML document: Reading an existing load XML
regular expressions to search for the tags and attributes, this file. It provides the createXMLReader function that returns an
will enhance the speed while reading XML contents. implementation of the XMLReader interface. It reads the root
element first and reads the sub element and corresponding
2 Fast XML Parser data. Finally creates the XML document and sends it to the user
To parse an XML document in software, the processing application as a XML document.
sequence starts by loading the XML document, then reading its
characters in sequence, extracting elements and attributes and Writing an XML document: Read the XML document and se-
then validating the XML document, writing parsed information parates the content and writes it to the corresponding hash
and finally reading the resulting parsed data. Our initial ap- table such as root element, sub element, attributes and data. It
proach separates the process of reading the XML document provides the CreateXMLWriter function to return an imple-
and stores the contents in to the hash table using regular ex- mentation of the XMLWriter function.
pressions. Fig.1 shows the architecture of the fast XML parser
tool. Knowledge based search: Search a particular element or con-
tent from an XML document by our fast XML Parser tool. In-
itially it checks the content in storage unit using hash key that
Application XML is root element. If it is available then it goes to the sub element
Parser and corresponding data and displays it as output. If it is not
available means it terminates the search.
XML <?xml version="1.0" encoding="UTF-8"?>

document 
<PRODUCTDATA>
<PRODUCT PROID="P001">
<PRODUCTNAME>Barbie
Doll</PRODUCTNAME>
Parsed <DESCRIPTION>This is a toy for children in
Storage Unit
Data the age group of 5-10 years</DESCRIPTION>
<PRICE>$120</PRICE>
<QUANTITY>12</QUANTITY>
</PRODUCT>
Fig.1: XML Parser Architecture </PRODUCTDATA>
The Fast XML Parser Tool contains four modules. First,

load an XML document in an application. Second, reading an Fig.2: Sample XML Document
XML document. Third, Writing an XML document into the ap-
plication. Finally, knowledge based search of an XML docu-
IJSER 2011
306

ISSN 2229-5518
3 EXPERIMENTAL RESULTS AND DISCUSSION

3.3 Parsing Performance
According to the experiment and considering the bal-
ance between memory and computing consumption, our pars- The results of parsing throughput per second are
er tool is appropriate for parsing XML document with less shown in figure. Note that our XML parser tool gets the lowest
memory. When XML documents size increases and gets larger, throughput, thus leading to low parsing performance. Figure 4
the time difference will increase accordingly. shows the cumulative parsing times for a group of similar XML
We made some assumptions to facilitate our estima- documents. The X-axis is the file size of the XML documents
tion. One of the most noticeable assumptions is that we simply that were parsed.
ignore the intensive garbage collection operations in DOM.
Garbage collection is actually very time-consuming, our parser
tool can avoid it, because we does not recycle the unused
memory space when parsing finishes. If we take this into ac-
count, the time difference would be even larger. So, we can
draw the conclusion that our parser gains great advantage over
DOM especially for large and complicated XML documents.
3.1. Testing Results
The testing result shows that great improvement has

been achieved over present XML parsers, both in terms of
memory computation and parsing performance. The tests are
performed on our XML parser tool and MSXML. MSXML is a
typical DOM parser which is widely used in enterprise applica-
tions. Similar tests are performed on other open-source XML Fig.4: Parsing Throughput Results
parsers too, like Xerces-C++, Jaxen and Xalan. For clarity, we
only present the testing results between our XML parser tool 3.4 Results Analysis
and MSXML. The testing codes are run in a HP G61 with
2.80GHz Intel(R) Core2 Due CPU and 4GB of RAM. From above testing results we can see that our XML parser tool
consistently outperforms MSXML in both memory consump-
tion and parsing performance. Our XML parser tool consumes
3.2. Memory Consumption memory about approximately the same size of original XML
document, while MSXML takes 3~4 times. Our XML parser tool
In this part, we test the ratio of memory space needed deliveries up to 20~35MB/sec sustained throughput, while the
to parse an XML document to the original XML document size. average parsing throughput of MSXML is only 12MB/sec. Note
The memory usage when many 1KB and 5KB XML documents that for MSXML, parsing throughput declines sharply when
are stored and accessed. The result shows in fig.3 that the XML documents become large, which verifies the conclusion
memory consumption is acceptable in such an environment. that too much creation and destroy of node objects in DOM can
slow down parsing performance considerably.
4 CONCLUSION
From a proper analysis of positive points and con-

straints on the XML Parser tool, it can be safely concluded that
the product is a highly efficient. This XML Parser tool is work-
ing properly and meeting to all user requirements. This tool
can be easily plugged in domain specific languages. Our Parser
tool can improved the XML parsing performance significantly
and scales well. The computing efficiency will be improved in
the future work.
5 REFERENCE
[1] eXtensible Markup Language, http:/www.w3.org/xml.

10th Feb 2011.
Fig.3: Memory Consumption Results
[2] Su Cheng Haw and G. S. V. Radha Krishna Rao, A Comparative
Study and Benchmarking on XML Parsers, International Confe-
rence on Computer science and Technology, IEEE Computer Society,
IJSER 2011
307

ISSN 2229-5518
pp.321-325, Feb-2007.
[3] Shiren Ye and Tat-Seng Chua, Learning Object Models from Se-
mistructured Web Documents, IEEE Transactions on Knowledge
and Data Engineering, pp. 334-339, 2006.
[4] Zhenghong Gao, Yinfei Pan, Ying Zhang and Kenneth Chiu, A
High Performance Schema-Specific XML Parser, Third IEEE In-
ternational Conference on e-Science and Grid Computing, IEEE Com-
puter Society, pp.245-252, 2007.
[5] A. Waheed, J. Ding, Benchmarking XML Based Application

Oriented Network infrastructure and Services, International
Symposium on Applications and the Internet, IEEE Computer Socie-
ty, 2007.
[6] Wei Zhang and Robert A. van Engelen, An Adaptive XML

Parser for Developing High-Performance Web Services, Fourth
IEEE International Conference on eScience, IEEE Computer Socie-
ty, 2008.
[7] Yunsong Zhang Lei Zhao* Jiwen Yang Liying Yu, NEM-XML:
A Fast Non-extractive XML Parsing Algorithm, Third Interna-
tional Conference on Multimedia and Ubiquitous Engineering, IEEE
Computer Society, 2009.
[8] Zhou Yanming and Qu Mingbin, A Run-time Adaptive and

Code-size Efficient XML Parser, 30th Annual International
Computer Software and Applications Conference, IEEE Com-
puter Society, 2006.
[9] Giuseppe Psaila, Loosely Coupling Java Algorithms and XML

Parsers: a Performance-Oriented Study, 22nd International Con-
ference on Data Engineering Workshops, IEEE Computer Society,
2006.
[10] Su-Cheng Haw and Chien-Sing Lee, INLAB2: Fast Native XML
Storage and Query Retrieval, 3rd International Conference on In-
telligent System and Knowledge Engineering, IEEE Computer So-
ciety, pp.44-49, 2008.
[11] Fadi El-Hassan and Dan Ionescu, SCBXP: An efficient hard-

ware-based XML parsing technique, IEEE Computer Society,
pp.45-50, 2009.
[12] Wei Lu , Kenneth Chiu and Yinfei Pan, A Parallel Approach to

XML Parsing, Grid Computing Conference , pp.223-230, 2006.
[13] Gang WANG, Cheng XU, Ying LI, Ying CHEN, Analyzing XML
Parser Memory Characteristics: Experiments towards Improv-
ing Web Services Performance, IEEE International Conference on
Web Services, IEEE Computer Society, 2006.
IJSER 2011

Journal April 2011 Edition

Uploaded by

Copyright:

Available Formats

Journal April 2011 Edition

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Journal April 2011 Edition

Uploaded by

Copyright:

Available Formats

International journal of scientific

and engineering research

Print: $50 per issue.

Reprints (minimum quantity 100 copies)

Copyright2011 IJSER Publishing, Inc.

1. Haptic Training Method Using a Nonlinear Joint Control

2. Prevalence of Eating Disorder among Female Students of Tonekaban University

3. Impact Of Leverage On Firms Investment Decision

4. Effect of Nanofluid Concentration on the Performance of Circular Heat Pipe

5. Enhancement of person identification using Iris pattern

6. Mn DOPED SnO2 Semiconducting Magnetic Thin Films Prepared by Spray

7. A novel comprehensive method for real time Video Motion Detection

8. Handoff Analysis for UMTS Environment

9. Performance and Emission Characteristics of Stationary CI Engine with Cardnol

10. Morphological Space and Transform System

11. Bus Proximity Indicator - (An Intelligent Bus Stop)

12. Designing aspects of Artificial Neural Network controller

13. Handwritten Character Recognition using Neural Network

14. Improved Performance of M-ary PPM in Different Free-Space Optical Channels

15. Electricity Sector Restructuring Experience of different Countries

16. Significant Role of Search Engine in Higher Education

17. Feature Selection for Cancer Classification: A Signal-to-noise Ratio Approach

18. Towards Analysis of Vulnerabilities and Measuring Security Level of Copmputer

19. Text Independent Speaker Identification In a Distant Talking Multi-Microphone

20. Novel Object Removal in Video using Patch Sparsity

21. New formula of Nuclear Force

23. Distributed Generation Planning Optimization Using Multiobjective

24. Marketing Of Asian Countries as Tourist Destination- Comparative Study of

25. Clinical and Computational Study of Geometry & Heamodynamics of Arterial

26. Segmentation Techniques for Iris Recognition System

27. Investigation of X-ray Plasmon Satellites in Rare Earth Compounds

28. Improvement of Power System Stability by Simultaneous AC-DC Power

29. A Comprehensive Distributed Generation Planning Optimization with Load

30. Positive Impact of Mobility Speed on Performance of AODV at Increased

31. Development and Physical, Chemical and Mechanical Characterization of

32. A chart for general and complicated data visualization

33. Application of Reliability Analysis: A Technical Survey

35. Design of Boost Circuit for Wind Generator

36. A Novel Method for Fingerprint Core Point Detection

37. Robust Impulse Eliminating Internal Model Control of Singular Systems

38. An Integrated Partial Backlogging Inventory Model having weibull demand

39. Survey of Current and Future Trends in Security in Wireless Networks

40. Anomalous hydrogen production during photolysis of NaHCO3 mixed water

42. Training and Analysis of a Neural Network Model Algorithm

43. Wireless Sensor Network: A Review on Data Aggregation

44. Slant Transformation As A Tool For Pre-processing In Image Processing

45. Separation of concerns in VoiceXML applications

46. Error of Approximation in Case of Definite Integrals

47. An Assessment model for Intelligence Competencies of Accounting

48. Processing of Images based on Segmentation Models for Extracting

Haptic Training Method Using a Nonlinear

I N the least decade the number of patients who have

damage can lea to impaired movement control upper 1.1 Justification

1.2 The Problem

and require continued rehabilitation.

1.3 Our Proposal 3.2 Haptic Device Dynamics

2 (5) Where c >>1, s s1, s 2,.., sn T denote dynamic error,

Where s denotes an error signal that we want to drive

Applying the antisymmetric property in (17), it reduces

For asymptotic convergence (Lyapunov stability theo-