POINT OF INTEREST AWARENESS USING INDOOR
POSITIONING WITH A MOBILE PHONE
Paulo Pombinho, Ana Paula Afonso and Maria Beatriz Carmo
Faculdade de Ciências da Universidade de Lisboa, Lisbon, Portugal
Keywords: Mobile Devices, Point of Interest, Mixed Environments, Indoor Positioning, Step Detection Algorithm.
Abstract: Although location based applications have been gaining popularity, most positioning devices do not work
when in an indoor environment, hindering the development of both mixed and indoor location based
applications. In this paper we present a point of interest aware application that shows information to the user
that is dependent on his indoor location. To be able to detect the user’s position, we propose a technique,
based on the detection of footsteps, and the direction in which they were taken, to be able to calculate the
position of the user inside a building. To improve the accuracy of the system, we use information about the
buildings floor plan to create a graph that can be used to correct invalid movements.
1 INTRODUCTION
Global positioning devices (GPS) are becoming
progressively more common in new mobile devices
and, for this reason, the real time information about
the location of users has become widely used in an
extensive range of location based applications, as
diverse as car navigation, city guides or geocaching
applications.
Indoor location based services can be used to
show relevant information about the users’ location
and help them navigate in an unknown building, or
track users with special needs, children, or inmates
inside a prison compound. The indoor location could
also be used to allow emergency services to explore
unknown areas in an easier and more efficient way
(Pahlavan, Li and Mäkelä, 2002; Jensen, Kruse and
Wendholt, 2009).
Despite being reliable and precise while in the
open, GPS devices need to be able to view a large
portion of the sky to be able to correctly calculate
the device’s position. This limitation has the
consequence of rendering the GPS useless while
indoors. Furthermore, alternative positioning
systems like the ones that use GSM or mobile phone
tracking do not have high enough accuracy to be
able to correctly identify a position inside a building.
Consequently, these restrictions hinder the
development of location based applications that also
focus on indoor use (Beauregard and Haas, 2006).
Our research is mainly focused on location based
services in mixed environments and the study of
visual adaptations to different contexts. We have
previously explored outdoor location based
applications, namely a point of interest visualization
application (Carmo, Afonso, Pombinho and Vaz,
2008) where information is presented over a map of
the region the user is currently at, and a location and
orientation based application that shows information
about points of interest that lie in the direction the
user is aiming the device at (Pombinho, Carmo,
Afonso and Aguiar). Hereafter, to allow the
exploration of indoor locations and enable us to
proceed our research in mixed environments, we
need to be able to obtain the position of the user
while inside a building.
Our goal is to develop an application that
enables the exploration of buildings unknown to the
user and also allow him to obtain information about
the different points of interest that may exist, inside
the building, near the position the user is currently
at. To achieve this, we first have to be able to
correctly estimate the user’s indoor position. To
avoid the limitations of positioning systems like the
GPS that, as referred, do not work indoors, or the
use of expensive and complex physical
infrastructures installed in each building, we present
an algorithm that allows the position of users inside
a building to be inferred from the movement done by
the user. To achieve this goal we use a mobile
device with an integrated accelerometer to detect
5
Pombinho P., Afonso A. and Carmo M..
POINT OF INTEREST AWARENESS USING INDOOR POSITIONING WITH A MOBILE PHONE.
DOI: 10.5220/0003351700050014
In Proceedings of the 1st International Conference on Pervasive and Embedded Computing and Communication Systems (PECCS-2011), pages 5-14
ISBN: 978-989-8425-48-5
Copyright
c
2011 SCITEPRESS (Science and Technology Publications, Lda.)
when the users takes a footstep, and a digital
compass to determine the direction of each footstep.
Using this information and the knowledge of the
buildings floor plan we can calculate with an
acceptable accuracy the indoor location of the user.
In the next section we will describe the current
indoor positioning technologies and techniques. In
section 3 our proposed positioning algorithms and
techniques are explained. In section 4, we present
the developed point of interest location aware
application prototype. In section 5, we present the
user experiments that were carried out. Finally, in
section 6 we conclude and point out future work.
2 INDOOR POSITIONING
TECHNOLOGIES
There are already some works that explore indoor
positioning mechanisms. Most of these works can be
divided in three main categories: techniques that use
physical infrastructures installed on the buildings,
explicitly for indoor positioning; techniques that use
existing Wi-Fi networks; and the techniques that use
sensors installed in the mobile device or the user
himself.
2.1 Infrastructures Installed for Indoor
Positioning
There are several diverse approaches that use
transmitters of some kind, installed on the buildings,
and corresponding receivers, carried by the user.
Hiyama et al. (2005) propose a museum guiding
systems using infrared (IR) transmitters installed at
each 1 meter interval on the ceiling of the Japanese
National Science Museum in Tokyo, and an IR
receiver embedded in a monaural headphone
connected to a museum guiding device. The infrared
transmitters cover overlapping areas, allowing the
receiver to recognize that the user is standing
between two specific transmitters.
Lim and Zhang (2006) propose a similar indoor
positioning method but, instead of IR transmitters,
use Radio Frequency Identification (RFID) tags
distributed around an area, enabling a user carrying
an RFID reader to get, at one time, more than one
RFID, and estimate his physical location. From the
experiments made, the authors conclude that this
approach is capable of an estimation error of only 1
meter.
Ghiani, Paternó, Santoro and Spano (2009)
implemented a mobile guide in the Carrara Marble
Museum and in the Calci Museum of Natural
History, where active RFID Beacon tags are placed
near the museum exhibits. The RFID reader carried
by the user allows the application to identify which
beacons are visible and, by analyzing their RSSI
(Received Signal Strength Indication) calculate
which of these the nearest one is.
Xu et al. (2009) propose a two step tracking
procedure. In the first step, a beacon-correlation
algorithm uses RSSI information from all the
detected beacon nodes to identify a wider area where
the user is. In the next step, a shadowing-grid
localization algorithm, using only the RSSI
information from the beacons present in the
identified area, is used to identify the accurate
location of the user.
Ikeda et al. (2008) propose the use of a wireless
sensor network using low power VHF radio to
provide indoor positioning services. This system was
implemented on the Yokohama Landmark Plaza in
Japan and uses a Smartphone that integrates a node
that receives beacon signals and processes them to
detect the user’s position and trajectories.
The PhoneGuide system (Bruns, Brombach,
Zeidler and Bimber, 2007) is a museum guidance
system that uses Bluetooth emitters in each exhibit,
which are identified by the user’s receiver. This does
not allow sufficient accuracy to differentiate
individual objects located within the signal range.
However it does allow the system to know which
exhibits are currently visible and, with this
information, to use the pictures captured by the
camera and compare them to the stored pictures of
the currently visible objects, to be able to recognize
to which one the user is looking at.
Rainer Mautz (2009) studied some recent indoor
positioning systems that are capable of achieving
cm-level accuracy. In his paper, he explores: the
iGPS system (Krautschneider, 2006) that uses a
transmitter of rotating fan-shaped infrared laser
planes that is capable of a mm-level accuracy; the
Locata System (Barnes, Rizos, Kanli and Pahwa,
2005) which is a positioning system that uses radio
signals transmitted in the 2.4 GHz band, to be able
to penetrate the inside of buildings, by transceivers
placed within a range of several hundred meters up
to several kilometres and with a sub cm precision;
and the Cricket (Priyantha, 2006), Active Bat (Hazas
and Hopper, 2006), and Dolphin system (Fukuju,
Minami, Morikawa and Aoyama, 2003), which are
ultrasonic positioning systems that have beacons
emitting both ultrasonic pulses and radio frequency
messages, and receivers that can use the difference
in the arrival of both the signals to calculate de
PECCS 2011 - International Conference on Pervasive and Embedded Computing and Communication Systems
6
distance from each beacon and triangulate their
position.
Kalkusch et al. (2002) present a mobile
augmented reality application where the user has a
helmet with a mounted camera that detects and
tracks well known markers, previously scattered
around the environment. This tracking allows the
system to identify the position and orientation of the
user.
2.2 Indoor Positioning using Existing
Wi-Fi Networks
Several systems have explored the use of Wi-Fi
wireless network access points.
Bahl and Padmanabhan (2000) developed the
RADAR system, which operates by identifying and
processing the signal strength information of
multiple base stations that are positioned in a way to
provide overlapping coverage, and triangulate the
position of the user with an accuracy of a few
meters.
Kitasuka, Nakanishi and Fukuda (2003) propose
the WiPS system. This system is similar to the
previous one, however, instead of only using
information from the wireless LAN access points,
the WiPS system also communicates with the
neighbouring WiPS terminals and uses information
obtained from them to be able to increase its
accuracy.
There are also commercial systems that explore
Wi-Fi Networks to provide indoor positioning. The
Ekahau Real Time Location System (RLTS)
(Ekahau, 2010) uses the existing Wi-Fi Networks to
provide sub room, room, floor or building level
accuracy, depending on the number of access points
available. Nokia are also currently developing an
experimental indoor positioning service at the
Kamppi Shopping Center in Helsinki, Finland, that
uses Wireless LAN (Nokia, 2009).
2.3 Infrastructure Free Indoor
Positioning
Kourogi et al. (2009) use an approach where sensors
are placed on the user’s waist, to detect walking
stance and velocity. This approach is then enhanced
by the use of surveillance cameras to estimate the
walking parameters.
Stéphane Beauregard (2008) focuses on the
indoor positioning of emergency first responders.
Shoe mounted sensors are used and a foot-inertial
pedestrian dead reckoning approach is proposed to
detect the displacement made by the foot in each
footstep and consequently the total displacement
made by the user. The placement of the sensor on
the shoes instead of other body areas allows better
results to be achieved when in the presence of
irregular walking patterns or sharp turns typical of
rescue missions.
Jiménez, Seco, Prieto and Guevara (2010) also
use shoe mounted sensors and developed the
INS-EKF-ZUPT framework to detect the position
and attitude of a person while walking. They also
employ a stance detection algorithm and several
methods for heading drift reduction.
Commercially, the Nike + iPod Pedometer
System (Nike, 2010) that has sensors placed in the
user’s shoes allows him to hear useful information
through his iPod about the exercise he is currently
doing.
Glanzer, Bernoulli, Wiessflecker and Walder
(2009) present an inertial pedestrian navigation
system that uses a low-cost micro-electro-
mechanical-system (MEMS) based Inertial
Measurement Unit consisting of a three-axis
accelerometer, a three-axis gyroscope, a three-axis
magnetometer, a temperature sensor, an A/D
converter and a microcontroller. From the data
obtained from these sensors, an integration of the
acceleration values is done to obtain changes in
position and attitude, and calculate the final position
of the user.
Köhler et al. (2007) have developed a prototype
for indoor positioning in small rooms, that uses a
compass, camera and grid projector. The system
projects a grid pattern into the environment to try
and determine the location of the planes
(corresponding to the walls) and intersections
(corresponding to the corners). If the system can
detect three orthogonal planes, for example two
walls and a ceiling, it can find its position and
orientation relative to that particular corner and, by
using a compass, find which particular corner it is
and consequently the relative position inside the
building.
Randell, Djiallis and Muller (2003) compare
different pedestrian dead reckoning techniques and
the use of different sensors that can be used to get
accurate measurements. They explored four different
case studies and were able to obtain accuracies of
the same order as a basic GPS receiver.
Ladstätter, Luley, Almer and Paletta (2010) use a
sensor placed in the pocket of the user to detect
shifts in acceleration and use the Weinberg
Expression to adjust the thresholds dynamically.
They have determined that the accuracy of the
technique is reliable for walking speed higher than 1
m/s, but has too many false positives when used at
POINT OF INTEREST AWARENESS USING INDOOR POSITIONING WITH A MOBILE PHONE
7
slow speeds.
Very recently, some preliminary works have
started to appear that focus only on the use of
sensors integrated in the mobile devices. Serra,
Carboni and Marotto (2010) present the early
developments of an indoor navigation system that
uses the device’s accelerometer, compass, camera
and internet connectivity. Despite using only the
sensor of the device, the proposed application uses
datamatrix drawn in maps scattered in the buildings
to obtain the buildings map and the starting position.
Some experiments have also been made recently,
to do an early evaluation of the accuracy of these
types of techniques. Dekel and Schiller (2010)
present an ongoing series of tests that explore indoor
navigation with un-augmented smart phones. They
were able to obtain an accuracy of more than 90%
on the detection of the distance travelled.
3 INDOOR POSITIONING
Despite providing some solutions for the indoor
positioning problem, the works presented in the
previous section are all either based on the existence
of a physical infrastructure in each building, or the
need of external sensors that are placed, for
example, on the user’s shoes or waist. These sensors
are a potential limitation to the natural movement of
the user or the practicability of the system.
Furthermore, some of the presented systems require
very expensive equipments and others, although
using relatively cheap beacons, require the
installation of a large number of them to be able to
obtain good accuracy.
Since our aim is to develop mixed environment
(indoor and outdoor) adaptive visualization
applications, our goal is to develop an approach that
does not need external sensors, beyond those
integrated in the mobile device, and that can be used
in buildings with no physical infrastructure installed.
Furthermore, since the user is holding the mobile
device in his hand, we want an approach that does
not hinder the users hand movements, allowing him
to retain some of the freedom of movement they
usually have when using a mobile device.
In the next subsections we will describe our
proposed approach. It comprises step detection and
the identification of the user’s position in a floor
plane graph.
3.1 Step Detection Algorithm
We use a 3-axis accelerometer integrated in the
mobile phone to capture, in real time, the
accelerations that the device is being subjected to.
When in a standing / resting stance, the only
acceleration that is present in the mobile device is
gravity. If we consider the mobile device to be
perpendicular to the floor plane, the X and Z axis
would be measuring no acceleration, and the vertical
Y-axis would be measuring the gravity, with a value
of approximately -9.8 m.s
-2
. Since the users can
freely tilt and rotate the mobile device, we have no
prior knowledge of which accelerometer axis is, at a
given point, measuring the vertical acceleration. For
this reason, despite potentially adding some noise,
we have chosen to analyze the resulting acceleration
vector norm, instead of analyzing each axis
separately.
Figure 1: Shifts in acceleration vector norm while
walking, represented in a blue thin line. The red thicker
line shows a single step pattern.
When walking, the user will, not only, apply a
forward acceleration but also, with a greater
magnitude, alternatively apply a vertical upward
acceleration followed by a vertical downward one.
Figure 1 shows an example of the acceleration
changes due to the user walking three steps.
We have defined four parameters that are used to
detect a footstep:
a peak amplitude λp that represents the
minimum positive shift in acceleration caused
by a footstep,
a trough amplitude λt that represents the
minimum negative shift in acceleration caused
by a footstep,
a minimum time interval Δtmin that needs to
go by before completing the footstep pattern,
a maximum time interval Δtmax that cannot
be exceeded for the footstep pattern to be
detected.
All of these parameters can be changed inside
the application in real time. The description of the
step detection algorithm is as follows:
PECCS 2011 - International Conference on Pervasive and Embedded Computing and Communication Systems
8
Step 1: calculate acceleration norm: a
Step 2: if(a > λp)
start timer t
start pattern detection
Step 3: while(t < Δtmin)
if(pattern detected)
discard footstep
stop pattern detection
Step 4: while(t > Δtmin AND t < Δtmax)
if(pattern detected AND a < λt)
footstep detected
Step 5: if(t > Δtmax)
discard footstep
The step detection algorithm works by constantly
checking if the current acceleration is greater than
λ
p
. When that happens, the application starts to
check if the footstep pattern is occurring. If a sudden
shift in acceleration occurs in less that Δt
min
, we
assume that it was caused by another movement and
the step is discarded. If the pattern occurs in more
time than Δt
min
but less than Δt
max
, and it reaches λ
t
creating a pattern similar to the one shown
previously, a step is recorded along with the current
orientation, which is obtained from the digital
compass. If after Δt
max
we still have not detected the
footstep pattern, the step is discarded.
To be able to identify the location of the user
inside a building, we use the last known latitude and
longitude obtained from the GPS as the initial
location of the algorithm. Afterwards, as each step is
detected, we use the orientation obtained from the
device’s digital compass and the defined average
step length to calculate the displacement made. By
adding all the displacements we can calculate the
trajectory taken by the user inside the building and
also his current location.
3.2 Floor Plan Graph and Position
Correction
Although the algorithm described in the previous
section can give an accurate positioning when used
for short periods of time, it can accumulate a
significant amount of error when used for longer
periods of time. To minimize errors caused by steps
that are not detected (false negatives), steps
incorrectly detected (false positives), or errors in the
step size, we make corrections to the position of the
user using information about the building’s floor
plan.
We have chosen to represent the floor plan by
the use of graphs, inspired by the works done by
Stoffel, Schoder and Ohlbach (2008) and Yuan and
Schneider (2010), in which the geometry of the
buildings floor plan is used to define a graph of the
possible paths a user can take.
We have opted to divide the floor plan in
rectangular areas of different sizes, where transition
areas (for example, doors) have the smallest size,
and areas with no transitions (for example, long
corridors with no doors) have the bigger areas. Each
of these areas corresponds to a node in the graph.
Figure 2 shows, on the left, an original floor plan
with the considered areas in red, and in the right the
graph that was defined, with a node for each area.
Figure 2: Definition of the floor plan graph.
The use of the graph allows calculating, with
each step taken, the position of the user inside a
certain node area. If at any time, the calculated
position is outside the current node area, the system
will verify if there is a valid transition from the
current node to another node in the specified
direction. If there is a valid transition, the system
places the user on the next node and calculates his
location in the new node area. If there is no valid
transition in the recorded orientation (i.e. there is no
direct path from the current node to the new one),
the system searches for the nearest position where
the detected movement would be valid and corrects
the user position.
4 APPLICATION PROTOTYPE
Our goal is to develop a point of interest aware
application that can enable a user to roam inside an
unknown building (for example, a shopping mall or
a university campus) and quickly obtain useful
information. The application should aid the user in
finding specific places and also to easily explore the
building while obtaining information about the
points of interest he is near to.
To test and evaluate our approach, we have
developed a simple application prototype for a HTC
Smartphone. We use the Smartphone’s integrated
accelerometer to obtain, in real time, the acceleration
POINT OF INTEREST AWARENESS USING INDOOR POSITIONING WITH A MOBILE PHONE
9
the device is being subjected to, and the digital
compass, to obtain the orientation of the device.
To obtain information about the points of
interest, we use a SQL Server database that contains
data about the name, location and attributes of each
point of interest. However, to illustrate the proposed
concept, we have defined a simple floor plan
composed of only nine different points of interest.
Figure 3: Prototype showing the desired destination.
The developed application has two main display
modes. In the first mode, used to support a search
task, if the user wants to search for a specific point
of interest, he can select the “Select PoI” button
from the “Options” menu and then search for the
specific point of interest he wants to find (for
example a certain store). When the user selects a
point of interest, it will be highlighted in the
building plan. Additionally, the available
information about that point of interest will also be
shown in the upper part of the interface (Figure 3).
The second mode is used to support an
exploratory task. If the user does not select any point
of interest, the application detects what areas the
user is near to, and shows information about them in
the upper part of the interface (Figure 4).
In our prototype, the main visualization area, in
the bottom part of the interface, shows the path the
users have taken drawn over the building plan and
their current location marked by a blue circle
(Figure 3). An example of path detection is shown in
Figure 4. The position of the user and the path he
has already taken is drawn in blue over the floor
plan. If the user position is automatically corrected
by the application, due to an invalid detected
movement, the interface displays a thick red line to
highlight the correction made.
Figure 4: Prototype showing the path the user has taken.
5 USER EXPERIMENTS
To accurately calculate the position of the user
inside a building, it is essential that we use adequate
step detection parameters and also consider an
accurate step size. To be able to obtain these values,
we performed a set of experiments to allow us to
understand the differences between the walking
patterns and step size of a diverse set of users.
5.1 Procedure
At the start of the experiment, we asked each user
their name, weight and height. Seven users
performed our experiments, four were male and
three female. They had diverse heights and weights,
ranging from 160 cm to 180 cm and 65 kg to 81kg,
respectively.
Afterwards, each user was asked to hold the
mobile device in their hand and follow the
instructions that were presented on the device’s
screen. For each experiment, the user’s were given
indications on the speed at which they were
supposed to walk, and instructed to press a start
button, walk ten steps and press the stop button.
The users were also instructed to stay on the spot
where they stopped, to allow us to measure the
distance they had travelled.
Each user was asked to carry out six tasks, three
of them outdoors and the remaining three inside a
shopping mall.
The users were first asked to walk in their
PECCS 2011 - International Conference on Pervasive and Embedded Computing and Communication Systems
10
normal walking speed when outdoors, then they
were asked to walk slowly while observing the
surrounding environment and, finally, they were
asked to walk as if they were in a hurry to get
somewhere. These tasks were then repeated while
indoors, where the users were asked to walk in a
normal speed, slowly while watching the shop
windows, and quickly as if they were late.
Finally, we also conducted some experiments
with the device secured on the user’s waist to be
able to obtain a control set with the device placed in
a more stable position.
5.2 Results
After having concluded the experiments, the
accelerometer logs, recorded by the mobile device,
allowed us to create charts displaying the variation
of the acceleration vector norm over time for each of
the tasks conducted by each user.
Using these charts we were able to analyse the
walking pattern of each user and determine, for each
step, what the optimal detection patterns would be.
Regarding the λ
p
and λ
t
parameters, we wanted
to understand if the minimum λ
p
value and the
maximum λ
t
value were sufficiently different that
they could be used as default values to detect all the
steps. However, as can be seen in Figure 5, the
values obtained are in fact very similar, and as
consequence are not adequate for the algorithm.
Figure 5: λ
p
and λ
t
acceleration parameters.
As an alternative, to obtain a set of values that
can detect the majority of the steps, we can use the
average and standard deviation for each parameter.
As a result, the value of λ
p
can be calculated by
using the average (12.95) and subtracting the
standard deviation (1.71). Similarly, the value of λ
t
can be obtained by adding the standard deviation
(1.11) to the average (7.53).
Regarding the time interval of the steps
(Figure 6), the average time spent on a step (245 ms)
and the relatively high standard deviation (99 ms)
suggest that a similar approach can also be used to
define the Δ
tmin
as the subtraction of the standard
deviation from the average and Δ
tmax
as the sum.
Figure 6: Time spent taking a step.
The values obtained allow us to detect the
majority of the steps taken by the users, however, as
can be seen by the standard deviation in the previous
charts, there are significant differences in the way
each user walks. For this reason, instead of using
fixed parameter values, it would be better for an
application to acquire information when the user is
outside (where his movement can be detected by the
GPS) to automatically calibrate the parameters.
Figure 7: λ
p
, λ
t
and Δ
t
indoor / outdoor ratios.
However, since the users do not move exactly
the same when indoors as they do while on the
outside, we also need to identify the ratio between
the two types of movements. To achieve this, we
calculated the ratio, for each user, between the
indoor and outdoor experiments at the same speed.
The average ratios are shown in Figure 7.
As important as detecting each step correctly, we
also have to choose the right step size since, in the
long run, it can originate a high amount of error. The
average step size obtained from our experiments is
65 cm. However, there are significant variations
depending on the speed of the movement and also if
it is indoors or outdoors (Figure 8).
POINT OF INTEREST AWARENESS USING INDOOR POSITIONING WITH A MOBILE PHONE
11
Figure 8: Average step size.
Despite the very diverse step lengths, when we
consider the ratio between each pair of indoor /
outdoor experiments, we obtain a ratio value that is
fairly constant (0.9) and that is directly proportional
to the speed of the user (Figure 9).
Figure 9: Step size indoor / outdoor ratio.
We also wanted to find the relation between the
weight, height and step size of the user. To obtain
the ratio, we calculated the (Height / Weight) / (Step
Size). The average ratio calculated is 0.039 and is
inversely proportional to the speed of the user
(Figure 10).
Figure 10: Height, weight and step size ratio.
Through the use of both the previous ratios and
information obtained from the GPS, when the user is
outdoors, it is possible to calculate with a low error
the average step length for each particular user.
5.3 Evaluation
To understand if the proposed approach was valid,
we conducted a small evaluation phase. This phase
had the goal of determining the accuracy of the step
detection algorithm.
Using information from the previous
experiments we calculated, for each user, their
optimal step detection parameters. After calibrating
the application, we asked the users to walk 50 steps
holding the mobile device in their hand and when
they stopped recorded how many steps the
application had detected. Each user conducted this
evaluation three times.
On average, the algorithm detected 47 steps of
the 50 (standard deviation = 2.5). Although the
number of steps detected can, in the long run, be
responsible for a high error, we believe that it can be
minimized through the use of the floor plan graph
position correction.
6 CONCLUSIONS AND FUTURE
WORK
In this paper, we have presented a point of interest
aware application that uses the information about the
user’s indoor location to display information about
the points of interest that exist in the vicinity of the
user. To be able to correctly identify the position of
the user, we have proposed an indoor positioning
method that does not need previously installed
physical infrastructures in a building. Furthermore,
this approach does not need external sensors,
avoiding the restriction of the user’s natural
movements when using a mobile device and walking
indoors.
We have performed some user experiments that
allowed us obtain valuable data about the differences
in the step patterns originated from a diverse set of
users. This knowledge can give us the insight on
what the best default step detection parameter values
are, and also help us in implementing an automatic
calibration of these values. A particularly important
parameter, which can cause a high accumulated
error, and should thus be automatically adjusted, is
the step size, since it varies not only between
different users, but also depends on the speed and
way the user is moving. Since our aim is to use this
algorithm in mixed environment visualization
PECCS 2011 - International Conference on Pervasive and Embedded Computing and Communication Systems
12
applications, one solution is to use information from
the GPS when the user is outdoors.
Although there is still work to be done to be able
to use this techniques, in a way they can be accurate
with a diverse group of users, and in complex
environments, the tests that we have done so far
indicate that it is a valid approach.
We intend, in the future, to more precisely
evaluate the precision of the algorithm, taking into
account the position correction using the floor plan
graph. We also intend to test more complex
buildings floor plans, and consequently the detection
of footsteps when going up and down a set of stairs,
and the use of escalators and elevators.
ACKNOWLEDGEMENTS
The work presented here is based on research funded
by the FCT - Fundação para a Ciência e Tecnologia
(Portuguese Science and Technology Foundation)
through the SFRH/BD/46546/2008 scholarship.
REFERENCES
Pahlavan, K., Li, X., and Mäkelä, J-P., 2002. Indoor
Geolocation Science and Technology. In IEEE
Communications Magazine (February 2002), 112-118.
Jensen, B., Kruse, R., and Wendholt, B., 2009.
Application of Indoor Navigation Technologies under
Practical Conditions. In Proceedings of the 6
th
Workshop on Positioning, Navigation and
Communication (WPNC’09). 267-273.
Beauregard, S., and Haas, H., 2006. Pedestrian Dead
Reckoning: A Basis for Personal Positioning. In
Proceedings of the 3
rd
Workshop on Positioning,
Navigation and Communication (WPNC’06). 27-36.
Carmo, M. B., Afonso, A. P., Pombinho, P., and Vaz, A.,
2008. MoViSys – A Visualization System for Mobile
Devices. In Proceedings of the Visual 2008 – 10
th
International Conference on Visual Information
Systems. 167-178.
Pombinho, P., Carmo, M. B., Afonso, A. P., and Aguiar,
H., 2010. Location and Orientation Based Queries on
Mobile Environments. Proceedings of the Computer
Graphics, Visualization, Computer Vision and Image
Processing 2010 International Conference.
Hiyama, A., Yamashita, J., Kuzuoka, H., Hirota, K., and
Hirose, M., 2005. Position Tracking Using Infra-Red
Signals for Museum Guiding System. In Proceedings
of the Ubiquitous Computing Systems, 2
nd
International Symposium. 49-61.
Lim, A., and Zhang, K., 2006. A Robust RFID-Based
Method for Precise Indoor Positioning. In M. Ali and
R. Dapoigny (Eds.): LNAI 4031, Springer-Verlag.
1189-1199.
Ghiani, G., Paternó, F., Santoro, C., and Spano, L. D.,
2009. A Location-Aware Guide Based on Active
RFID’s in Multi-Device Environments. In Computer-
Aided Design of User Interfaces VI, Springer London.
59-70.
Xu, H., Huang, L., Xu, T., Huo, Y., and Wang, Y., 2009.
Practical Indoor Tracking using Wireless Sensor
Networks. In Proceedings of the 6
th
Workshop on
Positioning, Navigation and Communication
(WPNC’09). 209-215.
Ikeda, T., Inoue, Y., Sashima, A., Yamamoto, K.,
Yamashita, T., and Kurumatani, K., 2008. ComPass
System: An Low Power Wireless Sensor Network
System and its Application to Indoor Positioning. In
Proceedings of the CSTST 2008. 656-662.
Bruns, E., Brombach, B., Zeidler, T., and Bimber, O.,
2007. Enabling Mobile Phones to Support Large-Scale
Museum Guidance. In IEEE Multimedia (April 2007),
Vol. 14 (2). 16-25.
Mautz, R., The Challenges of Indoor Environments and
Specification on Some Alternative Positioning
Systems. 2009. In Proceedings of the 6
th
Workshop on
Positioning, Navigation and Communication
(WPNC’09). 29-36.
Krautschneider, R., 2006. Untersuchungen zur
Leistungsfähigkeit des Messystems Indoor GPS.
Diploma Thesis at the University of Applied Sciences
of Karlshure.
Barnes, J., Rizos, C., Kanli, M., and Pahwa, A., 2005. A
High Accuracy Positioning using Locata’s Next
Generation Technology. In Proceedings of the 18
th
International Technology Meeting of the Satellite
Division of the U.S. Institute of Navigation. 2049-
2056.
Priyantha, N. B., 2006. The Cricket Indoor Location
System. PhD Thesis, Massachusetts Institute of
Technology.
Hazas, M., and Hopper, A., 2006. A Novel Broadband
Ultrasonic Location System for Improved Indoor
Positioning. In IEEE Transactions on Mobile
Computing, (May 2006), Vol. 5 (5).
Fukuju, Y., Minami, M., Morikawa, H., and Aoyama. T.,
2003. DOLPHIN: An Autonomous Indoor Positioning
System in Ubiquitous Computing Environment. In
Proceedings of the IEEE Workshop on Software
Technologies for Future Embedded Systems.
Kalkusch, M., Lidy, T., Knapp, M., Reitmayr, G.,
Kaufmann, H., and Schmalstieg, D., 2002. Structured
Visual Markers for Indoor Pathfinding. In Proceedings
of the 1
st
IEEE International Augmented Reality
Toolkit Workshop.
Bahl, P., and Padmanabhan, V. N., 2000. RADAR: An In-
Building RF-Based User Location and Tracking
System. In Proceedings of the IEEE Infocom 2000,
Vol. 2. 775-784.
Kitasuka, T., Nakanishi, T., and Fukuda, A., 2003.
Wireless LAN Based Indoor Positioning System WiPS
and its Simulation. In Proceedings of the IEEE
PACRIM’03. 272-275.
POINT OF INTEREST AWARENESS USING INDOOR POSITIONING WITH A MOBILE PHONE
13
Ekahau Wi-Fi Tracking Systems. http://www.ekahau.com
Nokia Kamppi Trial – Beta Labs.
http://betalabs.nokia.com/apps/nokia-kamppi-trial
Kourogi, M., Ishikawa, T., Kameda, Y., Ishikawa, J.,
Aoki, K., and Kurata, T., 2009. Pedestrian Dead
Reckoning and its Applications. In Proceedings of the
“Let’s Go Out” Workshop in Conjunction with
ISMAR’09.
Beauregard, S., 2008. Omnidirectional Pedestrian
Navigation for First Responders. In Proceedings of the
Workshop on Position, Navigation and
Communication (WPNC’08).
Jiménez, A. R., Seco, F., Prieto, J. C., and Guevara, J.,
2010. Indoor Pedestrian Navigation using and
INS/EKF Framework for Yaw Drift Reduction and a
Foot-Mounted IMU. In Proceedings of the 7
th
Workshop on Positioning, Navigation and
Communication (WPNC’10).
Nike + iPod – Meet Your New Personal Trainer.
http://www.apple.com/ipod/nike/
Glanzer, G., Bernoulli, T., Wiessflecker, T., and Walder,
U., 2009. Semi-autonomous Indoor Positioning using
MEMS-based Inertial Measurement Units and
Building Information. In Proceedings of the 6
th
Workshop on Positioning, Navigation and
Communication (WPNC’09).
Köhler, M., Patel, S. N., Summet, J. W., Stuntebeck, E. P.,
and Abowd, G. D., 2007. Tracksense_ Infrastructure
Free Precise Indoor Positioning Using Projected
Patterns. In A. LaMarca et al. (Eds.): Pervasive 2007,
LNCS 4480, Springer-Verlag. 334-350.
Randell, C., Djiallis, C., and Muller, H., 2003. Personal
Position Measurement using Dead Reckoning. In
Proceedings of the 7
th
International Symposium on
Wearable Computers.
Serra, A., Carboni, D., and Marotto, V., 2010. Indoor
Pedestrian Navigation System Using a Modern
Smartphone. In Proceedings of the MobileHCI 2010.
Dekel, A., and Schiller, E., 2010. DRec: Exploring Indoor
Navigation with an Un-Augmented Smart Phone. In
Proceedings of the MobileHCI 2010.
Ladstätter, S., Luley, P., Almer, A., and Paletta, L., 2010.
Multisensor Data Fusion for High Accuracy
Positioning on Mobile Phones. In Proceedings of the
MobileHCI 2010.
Stoffel, E-P., Schoder, K., and Ohlbach, H. J., 2008.
Applying Hierarchical Graphs to Pedestrian Indoor
Navigation. In Proceedings of the ACM GIS’08.
Yuan, W., and Schneider, M., 2010. iNav: An Indoor
Navigation Model Supporting Legth-Dependent
Optimal Routing. In Proceedings of the AGILE 2010.
PECCS 2011 - International Conference on Pervasive and Embedded Computing and Communication Systems
14