See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/319242767

An Android Based Medication Reminder System Based on OCR Using ANN

Article · February 2013

CITATIONS READS

3 4,467

1 author:

Parag Achaliya
SNJB's Late Sau Kantabai Bhavarlalji Jain College of Engineering
10 PUBLICATIONS   8 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Reoccurrence Based Webpage Segmentation View project

Data Security Using Cryptography, Steganography and LAN Messaging View project

All content following this page was uploaded by Parag Achaliya on 24 August 2017.

The user has requested enhancement of the downloaded file.

 International Journal of Computer Applications (0975 – 8887)
International Conference on Recent Trends in engineering & Technology - 2013(ICRTET'2013)

An Android based Medication Reminder System

based on OCR using ANN

Bhadane Ashwini Kale Sapna Bhuse Ishwari

UG Student, SNJB’s KBJ COE, UG Student, SNJB’s KBJ COE, UG Student, SNJB’s KBJ COE,
Chandwad, Chandwad, Chandwad,

P. N. Achaliya
Pawar Pallavi
UG Student, SNJB’s KBJ COE, Chandwad, PG Student, G. H. Raisoni Institute of Engineering &
Management, Jalgaon

include: a)irregular medicine in-takes due to the patient's busy

schedule, b) complicated in-take schedules due to the large
ABSTRACT number of medicines taken by the patient, c) adverse drug
Most of times patients may forget to take the medicines at proper reactions caused by un-reconciled prescriptions obtained from
time as per the specified in the prescription which may cause in different sources, d) lack of knowledge about proper use of
late recovery from the disease/illness. So it is necessary to take medicines[1]
proper medicines in proper quantity at proper time. In this paper
we introduce an Android based application for the patients. This In this paper we are introducing an Android application for the
application will remind their user to take proper medicines in patients which will remind their user to take proper medicines at
proper quantity at proper time by automatically setting the proper time by automatically setting the reminders in the mobile.
reminders in the mobile. These reminders will be automatically These reminders will be automatically set as per the prescription.
set by the application as per the prescription. This reminder will
remind their user patient that now it’s time to take the medicine.
2. RELATED WORK
Keywords- Android, medicine, prescription, disease/illness
Zao J.K., Mei-Ying Wang, Peihsuan Tsai, and Liu J.W.S.
1. INTRODUCTION discusses the approach of Wedjat – Smart Phone Application
which is based on to help patients to avoid medicine
Any living being can be a patient which may include human administration errors which are mentioned above.
beings, animals, pets, etc. The patients under human being
category may include businessman, social workers, politicians,
teachers, students, etc. These people may busy in their daily
routine life schedule. If they are suffering from any kind of
disease/illness then it’s their duty to take the proper medicines in
proper quantity at proper time. If the patient is at home then the
family members may remember & reminds patient to take the
medicines. But it is not possible for the family members to give
reminder by calling them when the patient person is out of
home/city. For this purpose there should be some facility for the
patients which will remind them about their medicine taken time.

Now days there are large number of mobile phone/smart phone Fig. 1 Wedjat - Smart Phone Application
users in the world. The bulky number of variety of applications
available in the mobile phone made the luxurious life. Mobile Wedjat can perform three primary functions:
phone companies are providing such a wonderful applications a) Issue medicine in-take reminders
for their users then question arises in mind that why not to use b) Provide medicine identification and in-take directions
those applications when company is providing them? Out of c) Maintain medicine in-take records[1]
those applications, Reminder facility in the mobile phone is the
most commonly used application which is used for preventing to Similarly, Prasad B discusses the approach of Medicine
remember each and every small thing. Reminder Pro which is based on reminding medicine schedules.
Most out-patient medication errors were made when patients
bought prescribed medicines from different drug stores and use
them at home without guidance. Common causes of these errors

25
 International Journal of Computer Applications (0975 – 8887)
International Conference on Recent Trends in engineering & Technology - 2013(ICRTET'2013)

c) User has to add the reminder manually about taking

the dose i.e. 2 times or 3 times in a day.
d) User has to manually select the week days about the
reminder.
e) Existing systems are not providing any facility about
end of medicines.
f) They are also not providing any kind of facility about
the original prescription.
g) There is nothing happening automatically. Everything
needs to do manually.

All these findings demands new application which will reduce

the manual work & do everything automatically. Also the
existing systems have some major drawbacks which motivate us
to develop new system. Those drawbacks are as follows:
Fig. 2 Medicine Reminder Pro B. Drawbacks of existing systems:
This free application supports up to 15 reminders. User can a) Reminders are not automatically sets. So every
select them in either repeating or non-repeating alarm patterns. existing system requires manual work of setting the
Any hourly time interval between alarms can be selected, reminder.
starting from the minimum of 1 hour. At the scheduled time, b) Existing systems are time consuming because of
application will produce a notification with an alarm, vibration manually setting the reminders.
or LED indication. [2] c) There is no facility of storing the original prescription
in any of the existing system.
Also Med Minder approach by David Garland of Garland d) There is no facility of reminding the doctor’s next
Systems, which as per their saying, is easy, free, full featured pill appointment in the existing system.
and medicine scheduler and medication reminder application. e) There is possibility of hanging down the existing
systems due to the manual work.

C. Table of Comparison
Table 1: Comparison of Existing System & Proposed System

Existing Proposed
Parameters
System System
Method Manual Automatic
More time Less time
Time
consuming consuming
Database Required Not required
Original prescription Not stored Stored
Reminder of
doctor’s next No Yes
appointment
Possibility of
hanging down an Yes No
application
Fig. 3 Med Minder

But this system has lots of manual work and it is much time 4. PROPOSED SYSTEM
consuming. [3]
The proposed system is an application for the Android platform
mobiles will remind their user about the medicine in-take
schedule. This reminder will be set in the mobile with the help of
the reminder application. The reminders will be automatically
set as per the prescription.
3. ANALYSIS
Android is a Linux-based operating system designed primarily
After studying and analyzing all the above existing popular for touch screen mobile devices such as smart phones and tablet
applications based on Android mobiles, some major findings computers, developed by Google in conjunction with the Open
noticed which reduce their popularity. Handset Alliance. Initially it was developed by Android Inc.
whom Google financially backed and later purchased in 2005.
A. Findings in existing systems: [4]
a) User has to add the name of the tablet/capsule
manually. No facility of adding them automatically. The proposed system will be developed for Android mobiles
b) User has to add the quantity/dose of the tablet/capsule only because the market share of Android is more than other
manually. No facility of adding them automatically. operating systems.

26
 International Journal of Computer Applications (0975 – 8887)
International Conference on Recent Trends in engineering & Technology - 2013(ICRTET'2013)

The output includes medicine in-take reminders (MIR) and

image of the prescription if required. After system produced an
in-take reminder, it inserts it into the Calendar maintained by the
mobile phone. The Calendar application then sets off an alarm at
the proper in-take time. If the patient forgets to take their
medicines then our system reminds them about medicine in-take
time.

6. MEDICINE SCHEDULING
SPECIFICATION
Table 2 shows the medication scheduling specification [1] [6] of
the specific medicine. MSS can be divided into three sections
Fig. 4 Comparison of Android & other OS in South-East containing prescription, dosage and interaction parameters
Asia Market [5] respectively.

Table 2: Sample Medication Scheduling Specification

A. Need of new system: Prescription Parameter (PP)
The existing systems are very useful; they are providing Medicine Identifier M
very nice facilities to the user about their medication reminding Medicine Dose g
& scheduler. But as we see in analysis section, the existing Medicine Form Capsule/Tablet/…
systems were developed with lots of major observed drawbacks Medicine Amount n
which need to overcome by developing new system. The Therapy Duration T
existing systems are not fully automatic; there are lots of manual Dosage Parameter (DP)
works like adding tablet/capsule name, dose per day, adding Min. & Max. Dose [dmin, dmax]
weekdays. Also they are not providing any notification about Min. & Max. Separations [nsmin, nsmax]
doctor’s next appointment, also the end of dose/medicine. Max. Intake B over interval R (B, R)
So there is a need of developing such system which will try Min. Intake L over interval P (L, P)
to overcome all the drawbacks of the existing systems. So we are Interaction Parameter (IP) <List>
taking the initiative for developing such full-fledged system Interferer Identifier N
which will be completely automatic like the system which we
Min. separation from M to N minToInterferer
are proposing will automatically set reminder about medicine in-
Min. separation from N to M minFrInterferer
take. Currently we are not providing the name of any particular
medicine & its dose in reminder. We will not providing any
facility for adding the reminder on weekdays instead reminder A. Prescription Parameter (PP)
will be set automatically as per the requirement mention in the
PP contains the necessary information of a medicine
prescription. Our system will also automatically give notification
including its name M, the dosage size g and the amount n to be
about doctor’s next appointment and lots of other user friendly
taken each time (as multiples of the dosage size) and its therapy
and automatic facilities.
duration T during which the patient should take the medicine. A
picture of the medicine is also incorporated to assist patients
5. SYSTEM OVERVIEW identifying the medicine.

Figure 5 shows input and output data flows of the proposed B. Dosage Parameter (DP)
system. The input includes user's medicine prescriptions and
DP states the minimum and maximum dose sizes [dmin,
personal preferences. The mobile application obtains user's
dmax] and the minimum and maximum separations [nsmin,
medicine prescriptions by taking its pictures. For personal
nsmax] between two consecutive doses. It also specifies the
preferences, the user may block time periods such as sleeping
supply rate (B, R), which the maximum dosage B is taken within
hours, religion activities time, etc. Proposed system tries not to
any time interval R and the demand rate (L, P) and which the
insert in-take events into those time slots.
minimum dosage L is taken in any time interval p.

Input Output C. Interaction Parameter (IP)

Doctor’s Medicine A medicine N that interacts with another medicine M is
Prescriptio Intake stated as an interferer of M. The IP section of medicine M
n Reminders contains an entry for each of its interferers. Two constraints are
specified for each interferer: minTolnterferer that specifies the
User minimum separation from medicine M to interferer N and
preference Image of
doctor’s minFrlnterferer that specifies the minimum separation from
s & System
Configurati
Proposed prescriptio interferer N to medicine M. [1] [6]
n (if
on Applicatio required) 7. OPTICAL CHARACTER
n RECOGNITION
Fig. 5 Input/output Data Flow

27
 International Journal of Computer Applications (0975 – 8887)
International Conference on Recent Trends in engineering & Technology - 2013(ICRTET'2013)

Optical Character Recognition (OCR) refers to identifying

printed characters as digitally recognizable form (such as
ASCII). Traditionally, this was possible only through complex
algorithms that had very little tolerance to errors. These
algorithms required more computational power and processing
time. In spite of these drawbacks, it is extremely difficult to
code.

With the advancements in Neural Networks, pattern recognition

has had a huge leap. This reduces coding to minimal. There are
many algorithms based on ANN to achieve OCR. [7]

7.1 Artificial Neural Network: Fig. 7 Stages of HCR

A neural network (NN) is a task with variable and tunable 7.2.1 Pre-processing
parameters. It is an arrangement relating with weighted
interconnection between neurons. Generally they are non-linear Directly implementing ANN modeling, it is important that
scalar transformation but can also be linear scalar the images should be in good quality which enhances the image
transformation. The following figure shows an example of 1 and decreases noise and distortion. This helps in achieving
hidden layer neural network with 3 inputs. higher accurate results. Preprocessing stage must be
implemented in any OCR system.

7.2.1.1 Noise Reduction

X1 ∑ ∂ It helps to decrease the noise in the image by
X2 ∑ performing a local averaging operation. In order to achieve this,
a median or mean filter can be used to achieve minimal blurring.
Y
∑ ∂ 7.2.1.2 Binarization
X3
Then the process of binarization is implemented on the
Fig. 5 Feed-forward network with 3 inputs, 2 hidden neurons resulting image. Since the color information of the resulting
& 1 output neuron image is irrelevant, binarization gives us uniformity between the
samples. It also reduces the computational power as it has to deal
with only 2 colors.
Artificial Neural Networks (ANN) is a collection of simple &
especially interconnected cells. These cells are set in such a way 7.2.1.2 Pixel Uniformity
that each cell derives its input from one or more other cells. The
information in ANN is always stored in a number of parameters. This process reduces the width of similar pixels to
These parameters can be pre-set by operator the ANN with uniform pixel which is done with the help of edge detection by
examples of input & also together with desired output. [8] Sobel’s method. Pixel Uniformity process reduces redundancy
and makes the characters uniform. Similarly other pre-
processing techniques such as Skew detection/correction,
histogram matching, etc. could also be done.

7.2.2 Feature Extraction

Feature extraction in OCR mainly refers to the extraction
of each character from the image.

7.2.2.1 Segmentation
Segmentation refers to separation of each character
from others. It is done by drawing the smallest rectangle drawn
around the characters.

7.2.2.2 Scaling & Binary Matrix Formation

The image thus extracted is scaled into a 32x32 matrix
irrespective of the size. The image is scaled up/down to achieve
Fig. 6 Multi-layer Artificial Neural Network uniformity. Scaling has to be done carefully preserving the vital
features which is done by implementing a separate user defined
function. If a line passes through a pixel, then corresponding
7.2 Implementation: pixel will be given value one (1), otherwise it is taken as zero
The process of handwritten character recognition can be (0). In this method the numbers of inputs are 32x32=1024 which
divided into 3 stages viz. Pre-processing, Feature Extraction, and requires more time to learn and classify the characters.
then passing into ANN for training.
7.2.2.3 Space Detection
Since each character is segmented by finding the space
between them it is not possible to find the actual space between
Pre Feature ANN words. In order to detect space, before passing each character to
Processing Extraction Modeling

28

Noise Reduction Segmentation Trained

Character Set
 International Journal of Computer Applications (0975 – 8887)
International Conference on Recent Trends in engineering & Technology - 2013(ICRTET'2013)

the ANN, the difference between the x-coordinates of adjacent 8. CONCLUSION AND FUTURE SCOPE
characters are compared with a threshold value to see if it’s an
actual space. Threshold value is calculated locally. The spaces
Currently there are many medication reminder systems which
between each character is calculated and sorted in descending
are operable manually. Due to manual work, the available
order. The difference between these sorted values determines
system becomes more time consuming. So in the given work, an
which one is an actual space (space between words) and which
attempt has been made to implement fully automatic medication
one is not. The lowest value acts as threshold.
reminder system based on handwritten character recognition.
7.2.3 ANN Modeling This is achieved with the help of artificial neural network.
Neural network is very effective to decipher any character of any
The ANN model selected for HCR is Feed Forward Back language. The accuracy of character recognition is more
Propagation Network (FFBPN). The information is passed in important. So accuracy of characters needs to improve by adding
Forward direction and Error is propagated backwards. The probability to each character. For example, a character Q is very
optimal number of hidden layers and nodes has to be selected less easy to find because character Q is more often mistaken
based on simulation and training results. Typically the model with O in most of the OCR systems. The proposed system will
contains 1 Input layer, 1 hidden layer and 1 Output layer. Input only set the reminders in the built-in calendar application of the
layer consist of 1024 nodes (32x32). No. of Output and Hidden mobile. This reminder reminds user about their medicine in-take
nodes vary according to the situation. Output nodes depend on schedule. The system which we are implementing will also give
the separate characters that make up the document. the reminder about doctor’s next appointment. It will also tell the
user of the end of the medicines. The scheduled reminder will
7.2.3.1 Trained Character Set not suggest any kind of medicine, dose of medicine, etc.
The trained character set consists of all the individual
letters/characters which are used to create a simple handwritten In future, efforts can be made to improve the accuracy of the
document. Each handwritten character has its unique feature. character recognition. Also the facility of adding names & dose
And these unique features associated with the characters remains of the medicine will be included in the reminder.
same throughout the same document.
9. ACKNOWLEDGMENT
7.2.3.2 Analysis
Analysis of a standard 26 English letters & some First of all I would like to thank my Hon’ble Principal Dr. V. J.
commonly used special symbols are done and documented. This Gond, SNJB’s KBJ College of Engineering, Chandwad (Nashik)
can be used as starting point to reduce time. for the motivation and inspiration that triggered me for this
work.
Number of nodes for ANN can be fixed after creating the final
trained character set of all proper letter & special symbols. In No volume of words is enough to express my gratitude towards
this case, there are 1024 input nodes representing each pixel of Prof. P. R. Bhaladhare, Head of Information Technology
32x32 characters. And output pixels can be set accordingly. [7] Department, who has been very concerned and aided for all the
essential materials required for the preparation of this paper.
7.2.3.3 Result They had helped me to explore this vast topic in an organized
manner and provided me with all the ideas on how to work
The image of handwritten document is given to ANN
towards a research-oriented venture.
model to check the accuracy of character recognition using
ANN. The input to & output of ANN is shown as below,
I would also like to thank my colleagues who were always there
at the need of the hour and provided with all the help and
facilities, which I required every time. I am very much grateful
to my family for their valuable support and inspiration that
helped me in conceptualizing the project.

10. REFERENCES
Fig. 8 Handwritten Characters [1] Zao J.K., Mei-Ying Wang, Peihsuan Tsai, Liu J.W.S.,
"Smart Phone Based Medicine In-take Scheduler, Reminder
In order to recognize the characters, the trained character set was and Monitor", IEEE e-Health Networking Applications and
created from the same handwriting. Following figure shows the Services (Healthcom), 2010
implementation of preprocessing & feature extraction stages on
[2] https://play.google.com/store/apps/details?id=com.dambara
the above handwritten characters.
.medremindpro&feature=search_result
[3] https://play.google.com/store/apps/details?id=com.garland.
medminderfree&feature=search_result
[4] http://en.wikipedia.org/wiki/Android_%28operating_syste
m%29
[5] http://www.devworx.in/news/android-news/android-
defeats-all-others-in-south-east-asia-market-survey-
Fig. 9 Preprocessing & Feature Extraction of Handwritten 122301.html
Characters

29
 International Journal of Computer Applications (0975 – 8887)
International Conference on Recent Trends in engineering & Technology - 2013(ICRTET'2013)

[6] Hsiu, PC, Yeh HC, Tsai PH, Shih CS, Burkhardt DH, Kuo Journal of Computer Applications (0975 – 8887) Volume
TW, Liu JWS, Huang TY, "A General Model for 52– No.12, August 2012
Medication Scheduling," Technical Report TR-JIS-05- 008,
Institute of Information Science, Academia Sinica, Taiwan, [8] Miguel Po-Hsien Wu, “Handwritten Character
July 2005. Recognition”, a thesis submitted to university of
Queensland.
[7] V. Kalaichelvi, Ahammed Shamir Ali, “ Application of
Neural Networks in Character Recognition”, International

30

View publication stats