Indoor Localization and Tracking Using Posterior State Distribution and Fingerprint Interpolation
Indoor Localization and Tracking Using Posterior State Distribution and Fingerprint Interpolation
Indoor Localization and Tracking Using Posterior State Distribution and Fingerprint Interpolation
ISSN 2278-6856
Abstract
Indoor localization is one of the essential modules in mobile
wireless application. In this paper, the problem of location
estimation and tracking of mobile user in a fixed wireless
network is addressed. A strategy based on empirical map of
received signal strength with interpolation to reduce the
calibration effort for radio map creation is proposed. The
proposed system models the motion dynamics of a mobile user
as a hidden markov model in which each state corresponds to
geo-location then tracking the user location using the posterior
HMM state distribution. We present experimental results that
demonstrate the ability of our tracking system to estimate the
user location with a high degree of accuracy and less
calibration effort.
1. INTRODUCTION
Indoor Localization System (ILS) have become very
popular in the recent years. The Indoor Localization
System refers to the applications that rely on location
detection as to detect a products stored in a warehouse,
location detection of medical personnel or equipment in a
hospital, location detection of re-men in a building on re
[1], [14], detecting the police dogs location trained to nd
explosives in a building, searching for tagged maintenance
tools scattered all over a plant, safety and healthcare [2].
The Global Positioning Systems (GPS) [3] are the most
widely used tracking systems for civilian positioning
service, and can offer accuracy close to 10 meters.
However, GPS are outdoor only cannot provide good
accuracy in indoor environments since the satellite signals
are blocked by building obstructions. Many indoor
localization system have been proposed such as Ultrasound
[4], Infrared Ray [5], Radio Signal [6], Cellular network
[7], Computer Vision [8], PHY information [9], Bluetooth
technique [10]. Most of these systems are able to provide
accurate results; however, they depend on additional
hardware or large-scale infrastructures. Thus, such systems
are hard to be widely deployed due to signicant cost and
energy consumption and specic environment range
limitations [11]. The indoor Localization techniques rely
on different kind of measurements that get relative position
information between nodes as ,among others, the angle of
arrival (AoA ;an estimate of the relative angles between
Page 47
3. DATA COLLECTION
Received signal strength data processing in indoor
positioning systems can be divided into two main groups.
One group is based on wave propagation and relies on
computing distances between mobile devices and points
whose coordinates are known. These second group is based
on mapping by combination of signal strength
measurements and geographical coordinates, called a
signal strength (SS) map.
ISSN 2278-6856
(1)
Page 48
ISSN 2278-6856
each measurements
is the power (dBm) of the signal
that is transmitted by the
AP as received by the MC.
4. SYSTEM MODELING
The proposed model is divided into two phases: offline and
online phases. During the offline phase, the RSS readings
from the base station are collected by the mobile device at
known locations, then an interpolation process will be
applied using kernel regression to generate fingerprint
database with a finer resolution. During the online phase,
RSS measured by the moving user from all the base station
are compared to the offline interpolated fingerprint
database to estimate its current mobile location. The
moving user dynamics are modeled as a markovian
homogenous first order process and therefore allow the
HMM to estimate and track the moving user position as
shown in figure 2.
(2)
Where u is a number from 0 to 1 that represent the fraction
of the distance between
(
) and
(
) at
which
(
) lies.
Page 49
ISSN 2278-6856
(5)
Where the expectation in (3) is taken with respect to the
discrete variable
which corresponds to x(t) at grid
location
The probability density function (pdf) of each measurement is
Here
is the forward variable in the
HMM like the following equation
Page 50
ISSN 2278-6856
(a)
(b)
(a)
Figure 8 Examples of position estimation of the mobile
node refrence path (Solid line), discrete position estimation
(Dots line).
of the thick walls which is more realistic than figure 7 (b)
that use path loss propagation model that neglect the
presence of any obstacles present in the indoor enviroment.
Each anchor node is modeled as an isotropic source with
2.4 GHz frequency, 0 dBm transmition power and 2 m
antenna height. The receiver nodes cover the whole area
with 0.5 m grid size and 1.5 m height to receive the
received signal strength of the signal from the anchor
nodes. The same mobile pattern in [1] is built to evaluate
and compare the position estimation accuracy with our
proposed model as shown in figure 8.
Page 51
ISSN 2278-6856
Bayesian
Filtering
HMM
MLE
HMM
HMMI
Grid
Spacing
(m)
0.5
Estimation
Error
(m)
0.77
Complexity
0.5
0.5
1.0
1.0
0.14
0.25
0.53
0.43
High
High
Medium
Medium
High
WKNN
Mean
Median
2.56
2.35
2.33
2.16
2.10
1.67
7.14
6.51
2.44
2.32
HMMI
Mean Median
2.00
1.63
1.64
1.37
1.52
1.36
5.35
4.65
1.81
1.35
Page 52
6. CONCOLUSION
In this paper, the problem of localization and tracking in
indoor environment has been approached. The proposed
tracking system architecture considers a static access point
network and a moving client moving in it. In this spirit, a
software simulator RPS ray-tracing and a real
measurements RSS map of a monitored environment is
used from experimental measurements. The formalization
of the problem and the mathematical modeling have been
developed and discussed in detail within the estimation
theory framework. A novel approach based on HMMI has
been proposed to track location of moving terminals. We
examined the performance of the proposed HMMI tracking
algorithm with Bayesian filtering in [1] and MLE using
RPS and also with WKNN with an experimental
measurement. The proposed HMMI achieves a less
complexity on reducing calibration effort with higher
localization accuracy.
References
[1] G. Ortolan and M. Bertinato, Low-Density Wireless
Sensor Networks for Localization and Tracking in
Critical Environments, IEEE Trans. on Vehicle
Technology , VOL.59 , NO.6 , JULY 2010
[2] K.Takata, J.Ma, and B.O.Apduhan, Adangerous
location aware system for assisting kids safety care,
in Proceedings of ACM 20th International Conference
on
Advanced Information
Networking and
ISSN 2278-6856
ISSN 2278-6856
AUTHOR
Reda
A.
El-Khoribi,
Prof.,
Information
Technology Dept.,
Faculty
of
Computers
and
Information,
Cairo
University,
Egypt. PhD in Electronics and
Communications
Engineering,
Cairo University 1998. MSC in
Electronics and Communication Engineering, Faculty of
Engineering, Cairo University, 1993. BSC in Electronics
and
Communication
Engineering,
Faculty
of
Engineering,Cairo University, 1988. His research interests
include: Image processing and Computer vision, pattern
classification, signal processing, computer graphics and
machine intelligence.
Page 54