“Seamless Indoor Outdoor

Navigation for Smartphone Users”

A Project Work
Submitted in partial fulfillment of the requirement for the degree
of
MASTER OF COMPUTER APPLICATION (MCA)
Of
Jadavpur University
By
KUHELI MANNA
Class Roll No: 002010503046
Examination Roll No: MCA2360041
Registration No: 154254 of 2020-21
Under the supervision of
Dr. CHANDREYEE CHOWDHURY
Department of Computer Science & Engineering
Faculty of Engineering and Technology
Jadavpur University, Kolkata-700032
2023

i
 Faculty of Engineering and Technology
Jadavpur University
To whom it may concern
This is to clarify that the project entitled “Seamless Indoor Outdoor Navigation for Smartphone Users”
has been completed by Kuheli Manna. This work is carried out under the supervision of Dr. Chandreyee
Chowdhury in partial fulfillment of the requirements for the degree of Master of Computer Application of
the Department of Computer Science and Engineering, Jadavpur University, during the session 2022-2023.
The project report has been approved as it satisfies the academic requirements concerning the project work
prescribed for the said degree.

(Signature of Head of the Department) (Signature of Project Supervisor)

Prof. Nandini Mukherjee Dr. Chandreyee Chowdhury
Head of the Department Project Supervisor
Department of Computer Science and Department of Computer Science and
Engineering Engineering,
Jadavpur University, Kolkata-700032 Jadavpur University, Kolkata-700032

(Signature of Dean of the Faculty)

Prof. Ardhendu Ghoshal
Jadavpur University, Kolkata-700032

ii
 Faculty of Engineering and Technology
Jadavpur University

Certificate of Approval
This is to clarify that the project entitled “Seamless Indoor Outdoor Navigation for Smartphone Users”
has been completed by Kuheli Manna. This work is carried out under the supervision of Dr. Chandreyee
Chowdhury in partial fulfillment of the requirements for the degree of Master of Computer Application of
the Department of Computer Science and Engineering, Jadavpur University, during the session 2022-2023.
The project report has been approved as it satisfies the academic requirements concerning the project work
prescribed for the said degree.

(Signature of the Internal Examiner) (Signature of the External Examiner)

iii
 Faculty of Engineering and Technology
Jadavpur University

Declaration of Originality and Compliance with Academic Ethics

I hereby certify that I am the sole author of this project and that no part of this project has been published or
submitted for publication. I certify that, to the best of my knowledge, my project does not infringe upon
anyone’s copyright nor violate any proprietary rights and that any ideas, techniques, quotations, or any other
material from the work of other people included in my project, published or otherwise, are fully
acknowledged by the standard referencing practices. I declare that this is a true copy of my project, including
any final revisions, and that this project has not been submitted for a higher degree to any other University or
Institution.

(Signature of the Candidate)

Name – Kuheli Manna
Registration No. - 154254 of 2020-21
Examination Roll No: MCA2360041
Class Roll No. - 002010503046
Project Title- “Seamless Indoor Outdoor Navigation for Smartphone Users”

iv
 Acknowledgment
First, I would like to thank my project supervisor, Dr. Chandreyee Chowdhury of the Department of
Computer Science and Engineering of Jadavpur University. The door to her office was always open
whenever I ran into a trouble spot or had questions about my project.
I would also like to thank the experts who were involved in the validation survey for this project, Prof.
Nandini Mukherjee, the Head of The Department of Computer Science and Engineering, and all the faculty
members of the Department of Computer Science and Engineering.
Among the seniors, I am deeply grateful to Manjarini Mallik for her guidance and supervision throughout the
project.
Finally, I must express my very profound gratitude to my parents and also to my friends cum classmates for
providing me with unfailing support and continuous encouragement throughout my years of study and the
process of my project work and writing this project. This accomplishment would not be possible without
them.
Thank you

(Signature of the Candidate)

Name – Kuheli Manna
Registration No. - 154254 of 2020-21
Examination Roll No: MCA2360041
Class Roll No. - 002010503046
Department of Computer Science & Engineering
Jadavpur University

v
 Abstract
As mobile technology continues to evolve, the demand for context-aware applications has grown
substantially. One crucial aspect of context awareness is accurately detecting whether a user is indoors or
outdoors, as it allows personalized and location-based services. This project presents a novel approach for
indoor-outdoor detection using a fusion of Wi-Fi, GPS, and Android sensor data to enhance the accuracy and
reliability of context-aware mobile applications.
To ensure our desired realistic and ubiquitous principles, we offer a unique indoor-outdoor system comprised
of four primary modules: (1) GPS, (2) Wi-Fi, (3) Magnetism, and (4) light intensity. The GPS sensor on
Android devices provides precise outdoor location information, while the Wi-Fi sensor continuously scans
for nearby wireless networks. By analyzing the characteristics of Wi-Fi signals, the system can detect indoor
locations with high accuracy. The android sensors, including accelerometer, light, magnetometer, and
orientation sensors, enhance the accuracy of the detection system. Extensive real-world experiments were
conducted in various indoor and outdoor environments to evaluate the proposed approach. The results
demonstrate the system’s robustness and effectiveness, achieving accurate indoor-outdoor detection under
different scenarios.

vi
 CONTENTS
1. Introduction....................................................................................................1
1.1. Overview and Motivation.......................................................................1-2
1.2. Project Contributions.................................................................................2
1.3. Project Outline...........................................................................................3
2. Background and Related Work.................................................................4-5
3. Experiment Design.........................................................................................6
3.1. Overview....................................................................................................6
3.2. Task Definition...........................................................................................6
3.3. Data Collection and Application development.......................................7-8
3.4. IO-detector Module Description and Application development.............8-9
4. Proposed work..............................................................................................10
4.1. System overview.................................................................................10-11
4.2. Light Detector.....................................................................................11-13
4.3. Magnetism Detector............................................................................13-15
4.4. GPS detector.......................................................................................15-17
4.5. Wi-Fi AP Detector..............................................................................18-19
4.6. Overall estimation....................................................................................20
4.7. IO-Detection algorithm.........................................................21
5. Experiment result.........................................................................................22
5.1.1. Experimental Setup.........................................................................22
5.1.2. Performance....................................................................................23
5.1.2.1. Performance of Subdetectors................................................23
5.1.2.2. Performance of Aggregated IO-Detector........................24-25
6. Conclusion and Future Work.....................................................................26
References......................................................................................................27-28

vii
LIST OF ABBREVIATIONS AND SYMBOLS

AP Access Point
API Application Programming
Interface APK Android Application Package
File GPS Global Positioning System
GSM Global System for Mobile Communications
IO Indoor outdoor
NA Not Applicable
RF Radio Frequency
RFID Radio Frequency Identification
RSSI Received Signal Strength Identification
Lux: the SI Unit of illuminance and luminous emittance.
dBm Decibel-mill watts
BSSID Basic Service Set Identifier
SSID Service Set Identifier

viii
Chapter 1

Introduction
The past decade has witnessed the flourishing of Smartphone technology and its application for indoor-
outdoor localization. Smartphones have developed at a rapid speed in recent years, and they are becoming
crucial equipment in daily life. Since navigation has been an essential requirement of daily life, smartphones
are widely used for navigation purposes nowadays. A location-based service has to be consistent in
providing navigation in both indoor and outdoor locations. Accurate indoor and outdoor localization is in
high demand for improving human daily life. One of the most fundamental items of this contextual
information is whether the device is in an indoor or outdoor environment because it makes a significant
difference if a user is standing in front of a shopping mall or inside a shopping mall. For determining the
current position of a mobile device, location-aware mobile applications need some capabilities. In the
outdoor environment, smartphone positioning system depends on Global Navigation Satellite Systems
(GNSS) including the Global Positioning System (GPS). GPS is one of the most reliable positioning systems
when it comes to differentiating between indoor and outdoor environments. In an outdoor context, GPS has a
high rate of accuracy because it can easily reach most the satellites. GPS needs to have a direct line of sight
between satellite and phone antenna, a smartphone inside an indoor environment will have a harder time
discovering satellites. That’s why GPS is often not effective in indoor environments [1].
Accurate indoor localization methods rely on several localization techniques such as sensor networks, RFID,
WI-FI, ultra-wideband (UWB), Bluetooth, cellular signals, etc. Wi-Fi-based indoor positioning is one of the
most popular positioning systems that have lots of pre-installed Wi-Fi access points [2]. Besides, by using
Wi- Fi for indoor positioning there is no need for additional peripherals and almost all smartphones are
equipped with Wi-Fi features. From nearby WI-FI access points, we can obtain receive signal strength (RSS)
that is used to characterize the fingerprint. In comparison with the outdoor environment, the accuracy of Wi-
Fi-based technology can not satisfy the expectation of indoor users because there are lots of points of interest
(POIs). When GPS is not used, Wi-Fi is a very common system, but certain locations have a scattered Wi-Fi
signal. Thus Wi-Fi localization is a good indoor localization method; however, it cannot guarantee, that there
are enough Wi-Fi fingerprints for localization [3]. However, the number of Wi-Fi AP is growing, and the
characters associated with Wi-Fi AP represent resources that could be helpful in the future specifically
indoor- outdoor detection Modern smartphones are now provided with numerous variety of sensors. The
most prominent sensors that most smartphones have are an accelerometer, magnetometer, gyroscope, light
sensor, proximity, microphone, and camera. We can detect various environmental aspects Using this sensor
data [4]. These sensors detect the user movement, providing vital data about user steps, direction, and
orientation. By utilizing the data from these sensors, we can enhance the accuracy and reliability of indoor-
outdoor detection.
In this project, we propose the indoor/outdoor detector: Seamless Indoor-Outdoor Navigation for
Smartphone Users using Wi-Fi, GPS, and smartphone sensors with careful consideration of the user’s
Environment. We primarily make use of four sensing resources: GPS, magnetic field, light sensor, and Wi-
Fi. In this approach, we examine in depth the adequacy of the devices and techniques such as Wi-Fi Aps
RSSI, GPS Signal-to- Noise-ratio (SNR), the number of connected APs or satellites, Magnetism, and
illumination. Through 2 monthly experiments, we observe that the light intensity, GPS signal, WIFI signal,
and the intensity of the magnetic field all individually exhibit distinct patterns in the indoor and outdoor
environment. These patterns are feasible for accurate classification of the ambient environment. More

1
preciously, light signals display distinct patterns when they are captured inside and outside the building.
Similarly, the RSSI from a Wi-Fi AP

2
by a device changes dramatically from the outdoor to indoor environments and the intensity of the magnetic
field also varies significantly inside a building. In this observation, we combined four sensing components
and developed an extensible indoor/outdoor detection app.
1.1 Motivation:
Indoor-outdoor detection is a powerful mechanism that bridges the gap between our physical and digital
worlds, enhancing our interactions with technology and our environment. In the last two decades, many
approaches have been proposed for Indoor/Outdoor detection and most research has been focused on a
general method employing semi-supervised machine learning and using light intensity, cellular signal
strength, and sound intensity [5]. This IODetector is quite poor because the system is invariant to changes
in relevant factors like environments, weather conditions, seasons, latitude, and devices, which ultimately
hurts the accuracy of the IO detection. Motivated by the above observations, we propose a new approach to
IO detection which aims to show accurate indoor-outdoor detection by combining GPS, Wi-Fi, and
Android sensor data so that it can significantly improve the navigation experience for users. Since
Smartphones are widely available, most smartphones are accomplished with a wide range of sensors.
Therefore users already own smartphones with GPS, Wi-Fi, and various sensors, there is no need to
provide additional equipment to users. Hence our approach is seamless, Real-time, and cost-effective. As
is presented in this project, if a user is walking down a busy street then GPS will accurately track the user’s
location. But suddenly users step into a building this is where Wi-Fi comes to the rescue with a Light
sensor. The fusion of GPS, Wi-Fi, and Android sensors is key to a world of infinite possibilities where
navigation is seamless, where buildings and open spaces are seamlessly connected.
1.2 Contribution:
We have designed a seamless indoor-outdoor detection system by combining Wi-Fi, GPS, and Android
sensors, which have been divided into four components such as light detector, magnetism detector, Wi-Fi
RSSI detector, and GPS SNR detector. By evaluating the confidence levels from these four sensing units,
we intellectually aggregate their detection results and guarantee an accurate, optimized result. In Chapter 4,
we will describe the design details of each component. We are also storing different types of sensor data in
a CSV file in different scenarios, such as indoor, outdoor, indoor-to-outdoor, and outdoor-to-indoor. By
analyzing these data sets, we are fixing the confidence level of those four sensing components. Using this
technology, we can accurately determine a user’s situation and improve overall user satisfaction.
To summarize, the contributions of the study are as follows-
1. We implement an Android app named sensor-data-collection-application that collects sensor data
from devices, such as GPS, Wi-Fi, magnetometer, accelerometer, proximity, and ambient light
sensors, and conduct experiments to collect data samples in various real daily scenarios. A data set
containing a large number of labeled data samples is constructed.
2. We have also implemented another Android app named IO detection for indoor-outdoor detection
based on data sets from the sensor-data-collection application. This app detects the user’s
environment (indoor or outdoor) and opens another app for Localization based on this detection. If
the user found in an indoor area, an app for indoor localization is opened, and if the user is found
in an outdoor area, Google Maps is open for outdoor navigation.

3
1.3 Project Outline
The rest of the project is organized as follows:
Chapter 2 shows the background and related work done in localization, positioning, and environment
techniques.
Chapter 3 shows the Experiment design and describes the experiment and data collection procedure.
Chapter 4 shows our proposed work, logic, and workflow for detecting the user’s ambient environment.
Chapter 5 shows the Experiment result including system performance evaluation in detail.
Chapter 6 concludes the project and discusses the possible future directions through which we plan to
extend our work.

4
Chapter 2

Background and Related

Works
In this chapter, we start by explaining a wide research works that have represented an important implicit
background to our proposed work and end by discussing on most explicit related work.
2.1 GPS
Even though GPS is the sensor with the most energy consumption, it is still the most reliable sensor when it
comes to differentiating between indoor and outdoor environments. GPS signals are highly dependent on the
line-of-sight (LOS) paths between the device and GPS satellites. It is distinguished that GPS signals are poor
in indoor environments as the LOS paths of GPS signals are blocked [6]. In the outdoor context, GPS has a
high rate of accuracy because it can easily reach most satellites as the LOS paths of GPS signals are not
blocked. Based on these facts, the localization accuracy of GPS or the availability of GPS signals has been
exploited to detect whether a device is in an indoor or outdoor environment [7] [8].

Figure 1. Illustration of a 24 GPS satellite constellation in motion relative to the Erath’s rotation (Source: Wikipedia).

Nowadays almost everyone can own a GPS receiver because the cost of GPS technology has significantly
decreased over the years due to evolution in technology. Its intuitive nature and easy implementation make
GPS an attractive tool for many researchers. Using GPS SNR value as an indicator for indoor-outdoor
transition is proposed in [9]. GPS signals are usually available outdoors where the sky is directly visible, and
are often weak or unavailable indoors when the sky is obscured by ceiling and walls. Thus, the estimated
accuracy of GPS localization can be used to detect if a user is indoors [10, 6].
2.2 Wireless signals
According to Vathsangam et al., [11], one of the suitable and cost-effective candidate techniques is using
existing Wireless Received Signal Strength (RSS) --based indoor positioning methods. Despite the
robustness of the Wi-Fi localization method, there is a challenge in evaluating Wi-Fi signals because of their
variety from time to time. The Wi-Fi signals hardly can be stable hence there is many complexities in the
indoor environment and it is difficult to get precise and flexible signal from wave propagations [12]. Wi-Fi
location determination consists of two primary methods, signal strength propagation models and
fingerprinting techniques [2]. The fingerprinting technique represents a reliable way of getting accurate
position information inside Wi-Fi networks but its data acquisition stage makes it slow, static, and hard to
scale. WiFiBoost [13] used a machine learning meta-algorithm that combined an adequate ensemble of
simple classifiers to improve the overall performance.

5
2.3 Multiple sensors
The following sensors are chosen to identify the indoor-outdoor state that might be related to the
environmental exposure of the user.
2.3.1 Light Sensor:
A light sensor, also known as a photo sensor, is a sensor that measures the intensity of light in its
surrounding environment. Light sensors operate based on the principle of the photoelectric effect. Light
sensors can be exploited to automatically adjust the lighting in indoor spaces depending on the ambient light
level. By comparing the light intensity readings from the sensors, it is possible to detect indoor or outdoor
environments. In the daytime, the sun provides around 10,000-80,000lx when it’s shining fully [14, 15],
compared to between 300 lx to 750 lx emitted by indoor lighting [16], and between 0.27 lx to 1.0 lx provided
by the moon under a clear sky [14]. From this, it is effective for indoor-outdoor estimation both in the
daytime and nighttime.
2.3.2Accelerometer:
Accelerometer can record the motion patterns and dynamics that relate to various activities and
environments. Indoor and outdoor environments often have different motion patterns. For example, walking
or running outdoors may involve faster changes in direction and variation in the ground compared to indoor
movement [17].
2.3.3 Magnetometer:
The magnetometer plays a vital role in the indoor navigation system. It is designed to measure the Earth’s
magnetic field, which is useful in navigation applications. This sensor is sensitive to disturbances caused by
electronics, magnets, and metals, and hence, the magnetometer variance is a good indication of nearby
structures and electronics, which happen mostly in an indoor environment [18].
2.3.4 Proximity:
A proximity sensor is a device that can detect or presence of nearby objects and for this, it does not need any
physical contract [19]. By utilizing proximity sensors it is possible to develop a system that can detect
indoor- outdoor environments.
2.3.5 Aggregated Sensor
Depending on the types and capabilities of the sensor, there may be situations where a single sensor might
not be able to tackle all application scenarios. The data coming from multiple sensors such as
accelerometers, proximity, and light sensors, wireless receivers [6], and magnetometers were exploited for
IO detection. Moreover, indoor-outdoor determinations rely on continuous sensing information from GPS
sources and internal sensors like accelerometers and gyroscopes to get an efficient estimation of user position
and pedestrian tracking [20]. IODetector [5] combined data from three lightweight sensors (light, cell tower
signal strength, and magnetic sensors) to develop an extensible IO detection framework that did not require a
training phase. There are some cases where these IODetector components fail to detect the right IO state
because of the non-adaptive nature of their respective thresholds. As an alternative, [21] proposes an
approach to IO detection that is based on semi-supervised learning. Note that the supervised learning-based
classifier approach has the same fundamental limitation as IODetector in that a supervised learning-based
classifier model trained in one environment may not be accurate in other environments.

6
Chapter 3

Experiment Design
3.1 Overview
The existing methods cannot estimate indoor/outdoor with high accuracy. We aim to the improvement of the
precision of indoor/outdoor estimation by using GPS, Wi-Fi, and Android sensors.
3.2 Task Definition
To create an exact indoor-outdoor detection algorithm, we divided our work into two parts. The first one is
the data collection part which is information coming from physical sensors in the device that could be
influenced by the environment. For that, we have developed an Android app for data collection. In The
second part, we analyzed these data sets and we developed another Android app that can easily detect indoor
and outdoor based on our analysis. Let’s see in detail below.
3.3 Data Collection Application development
We have developed an Android application for data collection. This Android app is implemented using
Android Studio. This app needs to access multiple sensors on the smartphone and save the sensor readings to
a CSV file. The class we used was SensorManager where this class lets developers access the device’s
sensors and provides the primary API for managing all aspects of sensors enabling, disabling, and collecting
information. The collected data consist of the number of GPS satellites, the GPS signal-to-noise ratio, the
number of Wi-Fi networks around the user, the highest signal strength of the Wi-Fi networks around the
user, luminance, the magnetic flux density, proximity, angular velocity, and linear acceleration. To acquire
data from GPS, the class we used was LocationManager. Then, we register a GPS listener in the
GpsStatus.Listener method, and start handling incoming data in the onGpsStatusChanged() callback method
that can used for receiving notification when GPs status has changed. We picked Wi-Fi Manger for the Wi-
Fi scan because it provides the primary API for managing all elements of Wi-Fi connectivity, including a list
of configured networks, presently active Wi-Fi networks, and access point scans. To acquire data from other
sensors, we first need to create an instance of the SensorManager class, which we can use to get an instance
of a physical sensor. Then, we register a sensor listener in the onResume() method and start handling
incoming sensor data in the onSensorChanged() callback method. This application provides a user-friendly
interface that allows users to interact with the app. This app includes various features such as data
visualization, file manager settings, configuration settings, dark mode turn on turn off the feature, and
Google map visualization.
It is already maintained that this application is user-friendly so that the user can easily specify whether he/she
is indoors or outdoors. To start data collection user has to press the start button and also specify whether
he/she is in an indoor or outdoor environment Labeling would be done in this circumstance. Thus, the user
can modify data labels or remove the collected data if they make any mistakes. If the user wants to terminate
data collection, then he/she has to press the ‘save and exit’ button. This button saves the collected data to
their respective named CSV file. he file name format of CSV files is determined by the date and time. The
process of data collection can be 10 minutes, 30 minutes, or unlimited time depending upon the user. The
user can stop data collection at any time. This makes the process of starting and stopping the data collection
fast and simple for the user.

7
3.3.1 Data collection
In this data collection process, the users, carrying smartphones, walk around inside the building or outside
the building as usual for daily activities. Their smartphones record relevant information from various
positions. To evaluate our IO-Detector approach, using a custom Android app, we collected different
samples of sensor data with a fair distribution between two environments: the Jadavpur University campus
and a residential area. For each site, we collected light signal six times, magnetism signal four times, GPS
signal two times and Wi- Fi signal two times on average with different sampling rates. To assess the
accuracy of different approaches, the Android app relies on an interface for volunteers who participated in
the data collection to manually input (indoor/outdoor) ground-truth information. There were a total of three
members who participated in the data collection process and various models of smartphones are used for data
collection. The users collected the data in their daily lives in both urban and rural areas. This ensures the
diversity of the data set.

Fig. 2. Data logging process. The picture shows a user configuring the data logging session.

Table 1
Example of collected data instance.
" sensor_data ":
" Gps_lat ": 22.49569877,
" Gps_long ": 88.3725543,
" Gps_accuracy ": 7.62,
" Gps_snr_average": 27.53999,
" Gps_satellites ": 5,
" Gps_snr_Trend ": 4.2222233,
" wifi_AP_No ": 17,
" wifi_Max_RSSI ": -48,
" wifi_Avg_RSSI ": -85,
" 1st_AP": 1,
" 2nd_AP": 4,
" magnetic_x ": 32.156255,
" magnetic_y ": -36.487503,
" magnetic_z ": -15.000001,
" mag_veriance ": 0.7068694,
" light ": 303.155347,
" proximity ": 5.0,
" accelerometer_x ":-1.72905,
" accelerometer_y ": 4.02,
" accelerometer_z ": 8.95005
" Result": "outdoor",

8
3.3.2 Overview of the data set
The dataset contains a series of recorded readings that are captured at regular intervals by the sensors. For
example, magnetometer datasets may include 3-axis magnetic values(x, y, z) while light datasets may
provide light values. GPS datasets may include latitude, longitude, altitude, accuracy, and timestamp values.
Each sensor reading is normally correlated with a timestamp that describes the exact time when the
measurement was recorded. This dataset may indicate the frequency at which the sensor readings were
sampled. For example, Wi-Fi datasets normally have a lower frequency compared to accelerometer,
magnetometer, proximity, etc. This is because Wi-Fi scans, which involve measuring the RSS of nearby
access points, are typically performed every 10 to 30 seconds or even longer intervals. Higher data frequency
as a result in larger datasets requires more storage capacity.
3.4 IO-detector
3.4.1 Module Description
In this project, we introduced the IO-Detector which is a combination of “IO” (including indoor-outdoor)
and “detector”, signifying the capability to distinguish between indoor and outdoor environments. The
intuition behind our system is that simply performs real-time Android applications running on smartphones
that can be activated by any user when needed. The main goal of our system is to provide accurate indoor-
outdoor detection results. IO-detector classifies the surrounding environment into three main categories:
indoor, semi- outdoor, and outdoor, based on signatures observed from various physical sensors like light,
magnetism, GPS, and multiple Wi-Fi AP signals during user transitions from inside buildings into outdoor
and return. We conducted experiments in controlled areas where data was collected from fixed routing traces
that can provide enough description and enhance the certainty about the type of those surrounding
environments. IO-Detector mainly achieves many practical design requirements:
 High accuracy –as we are using Wi-Fi, GPS, light, and magnetic sensors. Therefore if any of these
module does not perform well then the other module configures the user's ambient environment. For
example, sometimes indoor GPS does not provide SNR value. Thus GPS is inactive. In this situation,
Wi-Fi can play an important role but if there is no Wi-Fi AP then Wi-Fi can’t find Wi-Fi fingerprint.
For that, there is only one way left which is indoor-outdoor detection using smartphone sensors. In
this project, we proposed two sensors Light and magnetometer. There may be a case where light is
not available but the magnetic field always works. Combining all four modules we can say this
approach is highly accurate.
 Universal applicability- IO-Detector has large flexibility and scalability to run proactively on all of
today’s smartphones and smartphones are widely available. That ensures its wide applicability.
 Real-time –IO-Detector continuously monitors and updates the indoor-outdoor status as a user or
device moves and the environment changes.

9
3.4.2 IO-detector Application Development
We have developed another Android application for Indoor-outdoor detection. This Android app is also
implemented using Android Studio. It applies to most brands of Android phones such as Samsung, Redmi,
and LG. This app needs to access location and Wi-Fi services on the smartphone for detecting the ambient
environment. Therefore this app needs permission for Location access. If the user isn’t allowed to permit
access to location service then this app will show a prompt and once again it ask the user to permit location
service. In the case of Wi-Fi, this app will ask the user to please turn on Wi-Fi. This application also stores
relevant data in a CSV file concerning detection results. After opening this app it shows nearby Wi-Fi lists
with their SSID, BSSID, Capabilities, level, and frequency. This list is constantly sorted by the RSSI
showing the access point with a higher level of signal strength at the top of the list. To start Wi-Fi scanning
user has to press the scan button. After the scan button is pressed relevant Wi-Fi list will be updated if there
is an update and whenever the scan button is pressed the user will be shown that scanning is in progress. It
gives the user a user-friendly atmosphere. For indoor-outdoor detection, the user has to press the start button
and press button to show the user whether it is outdoors or indoors. If the IO-Detector finds that the user’s
environment is indoors, then it will open an indoor localization app to help the user find his/her path. In the
case of outdoor it directly open Google map to help the user navigate. We will show in detail in the
implementation section the tasks that will be performed in the background of this application after the button
is pressed.

10
Chapter 4

Proposed work
In this section, we introduced the system architecture and design details for each component in IO-Detector.
Then we use a Light sensor, magnetic field sensor, Wi-Fi, and GPS to determine indoor and outdoor states.
Each of these four components works independently to estimate a state and a confidence value. The final
state is determined after adding all the confidence values. Figure 4.1 presents the system architecture of the
IO- Detector.

Figure 4.1 System architecture of IO-Detector.

4.1 System overview

In this project, we introduced an IO-detector, “Seamless Indoor Outdoor Navigation for Smartphone Users”
which has large flexibility and scalability to run actively on all of today’s smartphones. We aim for the
improvement of precision of the indoor/outdoor estimation by using four modules that are maintained in
System architecture. Fundamentally, the IO-Detector is built based on four experimental observations: (1) In
the daytime, outdoors, light intensity is typically above 2000 Lux; (2) When the user’s context changes from
outdoors to indoors, the Wi-Fi signal strength increases due to neighbor access points; and (3) Magnetic field
sensed by the phone tends to change rapidly when the user is moving indoors where there are possibly many
appliances, electric currents and metallic objects nearby, compared to outdoors. (4) In the outdoors, the SNR
of all GPS signals is more than 20dB and the GPS module of the terminal can communicate with more than 3
satellites. The indoor/outdoor estimations by the 4 physical quantities mentioned above are carried out in
parallel. In addition, the estimation reliability is maintained in each estimation. If each estimated result is
different, the estimated result with the highest reliability is adopted. In the rest of this section, we will
describe the design details of each component. To reveal the signal features with different environments, we
empirically study the patterns of the light signal, Wi-Fi signal, GPS SNR, and magnetism signal in different
environments. All of the signals are collected in different environments under different weather conditions,
including sunny, cloudy, and rainy days, and at different times of the day. The light signal and magnetism
are collected with
11
different orientations of the light sensor and magnetic sensor, Wi-Fi signal and GPS signal are collected
when the user walks from outdoors to indoors and vice versa.
4.2. Light Detector
The primary source of light for light detection is the Sun. Light Detector is designed to capture and measure
the intensity of light, and it can detect sunlight as well as other sources of light. In outdoor and semi-outdoor
environments, the sun emits electromagnetic radiation across the broad spectrum including visible light
which is the primary light source in the daytime. Indoors, artificial light sources play a crucial role in light
detection.
4.2.1 Light intensity
Light intensity refers to the amount of light energy present per unit area that is measured in units such as
lux, foot candles, or watts per square meter. We normally observe that during the daytime, the intensity of
light inside a building that relies on artificial light is typically much lower than in an outdoor or semi-
outdoor environment that is based on sunlight, even on cloudy or rainy days. The major reason is that the
intensity of sunlight within the visible spectrum is normally much higher than that from ordinary lighting
lamps. The luminous flux from sunlight is much higher than that from artificial light sources during the
daytime. Outdoors, there is direct exposure to sunlight, and there are typically no walls, ceilings, or other
structures blocking the light path, allowing light to spread more freely and reach larger areas. Therefore,
we can accurately distinguish the indoor environment from the outdoor or semi-outdoor environment by
using the observed light intensity.

35000 Light outdoor to indoor to outdoor

Light Intensity
30000
Light Intensity( Lux)

25000

20000
outdoor indoor outdoor
15000

10000

5000

0
Time(s) 0 50 100 150 200 250 300 350 400 450

Figure 4.2.1. Light intensity pattern when the user moves indoors, then returns out outdoors during a sunny day with a smartphone
in hand and screen facing up.

There may be Additional information is needed to make the light information more useful. First, the phone
might be in the pocket of the user, which would mean the light sensor is obstructed and no light
measurement is available. For this scenario, we included the proximity sensor, and the idea is to determine
when the information from the light sensor is reliable. We might often experience that when we are on a call,
the screen of our smartphones will turn off automatically when we bring the phone to our ears, and the
screen will turn on when we take it back. This is because the proximity sensor recognizes the object near the
phone. Secondly, the clouds can significantly affect the amount of light received from the sun. To address
this issue, information about the weather was included. Lastly, the sun is not available 24 hours a day. This
changes depending on the time and place the user is. To further improve the possibility of inferring
information from the light sensor, information about the part of the day was collected. Even though a simple
12
approach would be to divide the day

13
into 2, daytime and nighttime, Based on sunrise time and sunset time, it is decided whether it is daytime or
night. In our work, we exploit time clock sensors and measurements to determine current user time either in
the daytime or at night to enhance light intensity module performance as it will be later.

400
(a) Indoor Light (b) Outdoor Light
20000
Light Intensity (Lux)

Light Intensity (Lux)

300 15000
200 10000
100 5000
0 0 20 40 60 80 0
Time(Seconds) 0 10 20 3 0 40 50
Time (Secon d s
)

6000 (C)Indoor-Outdoor light

Intensity(Lux)

4000
Light

2000

0
0 5 10 Time(Sec)15 20 25 30

Fig. 4.2.2. Light intensity exhibits a periodic pattern in the indoor environment during the daytime as shown in (a), whereas the
Periodic pattern in outside buildings during the daytime is shown in (b). The overall trend of light intensity is shown in part (c).

In this experiment, the light sensor of the terminal measures the intensity of light every 1 second. The
intensity of light has a large difference in values between indoors and outdoors in the daytime and also at
night. In the daytime, the outdoor illumination is about 3000-80000 lux, which is shown in Fig 4.2.2(b). In
contrast, the indoor illumination is about 70-800 lux, which is shown in Fig 4.2.2(a). From Figure 4.2.2(b),
the light intensities in both outdoor and semi-outdoor scenarios are above 2,000Lux and much higher than
that in the indoor environment in the daytime (from 8:00 AM to 5:00 PM). In the nighttime, the outdoor
streetlight is about 1-5 lux, and the brightness of the full moon is about 0.01- 0.1 lux. In the night outdoors,
illumination is almost 0 lux shown in fig 4.2.3. The illumination of the general indoor fluorescent lamp is
about 20-800 lux. From this, it is thought that the illumination is effective for indoor/outdoor estimation both
in the daytime and the nighttime. In addition, the light intensities in the indoor and outdoor environments are
both relatively stable.
indoor to outdoor Light (night time)
300
Light Intensity

200 Light not available in outdoor at night

100
Indoor
0
0 20 40 60 Time(sec8) 0 100 120 140 160
Figure 4.2.3. Light intensity pattern when the user moves indoors, then returns outdoors during the nighttime with a smartphone in
hand and screen facing up.
14
 Figure 4.2.4: Light intensity sense module flowchart

Finally, based on those observations and challenges, we construct a light intensity module algorithm as
shown in algorithm 4.2.4. From the 4.2.4 flowchart, we try to get some information related to the light
intensity module like maximum and current light measurements, proximity (Far or near), and current system
time clock.
During experiments, we collect 24-hour light intensity readings to monitor light signature variations in
indoor/outdoor environments all day by using discrete readings from mobile devices only. Even when the
light sensor is back to the sun, the light intensity is relatively high as well (e.g., around 3,000Lux). Figure
4.2.1 indicates that outdoor illumination becomes higher than indoor illumination in the daytime and figure
4.2.3 indicates that indoor illumination becomes higher than outdoor illumination in the night. Before and
after sunrise and sunset respectively, no light from the sun is perceivable in the atmosphere, which would
mean that any light perceived by the light sensor is artificial light or moonlight. After a few minutes of
sunrise, the outdoor illumination gradually increases from 0 lux to 10000 lux. Lighting estimation has great
credibility between 8:00 AM and 4:00 PM. The estimation by brightness is unreliable near dawn and
nightfall Because of the influence of the buildings' shadows, the terminal's orientation, and clouds, the
exterior light swings considerably from 10000 lux to 80000 lux at this time. The major reason behind this is
that these periods represent critical times when outdoor environment brightness is much less. The limitation
of the light detector is that the light signal is not always available. Not only that using light sensors we
cannot confidently distinguish the outdoor and semi-outdoor environments.
4.3 Magnetism Detector
Magnetic field is an invisible for created by magnetic objects or electric currents that surround us, shaping
our environment and influencing countless aspects of our daily lives. It possesses both strength and
direction, and its effects can be observed through the behavior of magnetic materials and the interaction
between charged particles. Earth itself has a magnetic field which play a crucial role in various aspect. The
disturbance of the earth’s magnetic field inside buildings can be utilized as fingerprints for indoor
localization. The quantity of magnetism is easy to use for the estimation because there is a large change in
the quantity of magnetism indoors. Though the previous setting including the measurement in the building
is necessary to use magnetism for indoor positioning estimation, only the observation of the change in the
quantity of magnetism is necessary for the indoor/outdoor estimation.
15
 (a) Manetometer variance
60
outdoor
Magnetic Variance

40 indoor

20

0
0 5 10 15 20 Time25(sec 30 35 40 45 50
)
(b) Ambient magnetic field
60
Magnetic field

40
intensity

indoor magnetism average 50

20
outdoor magnetism average 40

0
50 0 100 150 250 300 350 400
20T0ime
Fig. 4.3.1 The variation of magnetic field intensity. The magnetism signal varies significantly when the user moves to indoor
environments but keeps relatively stable when the user gets to outdoor environments.

Figure 4.3.2: Flowchart of estimation by magnetism

The magnetometer variance is a good indication of nearby structures and electronics, which happen mostly
in an indoor environment. Figure 4.3.1(a) plots the magnetic field intensity and its variance in two different
scenarios indoor and outdoor. The magnetism sensor of the terminal measures the quantity of magnetism
every 1 second and calculates the variance of the magnetism every 10 seconds. We find that the intensity of
the magnetic field in the indoor environment varies dramatically. In indoors, we find that the variance is very
high when the user moves (from the 1st second to 23th second). When the user is walking through the
corridor, the magnetic field intensity also shows significant variance (from 23 seconds to 30 seconds).In
outdoors, If the variance of magnetism is less than the threshold, it is estimated to be outdoor. And if more
than or equal to the threshold, it is estimated to be indoor. The threshold of the variance is set to 18. Fig.
4.3.2 shows the flowchart of the indoor/outdoor estimation using a Magnetometer.

16
The estimation by magnetism cannot provide accurate results if the terminal is covered in a bag or a pocket.
Therefore the covered situation of the terminal is estimated by a proximity sensor. The magnetic variance is
below the threshold for some cases, both indoors and outdoors. In these cases, the IO-Detector component
magnetism detector fails to detect the right IO state.

4.4 GPS detector

A "constellation" of 31 evenly spaced satellites known as the GPS (Global Positioning System) orbits the
Earth and enables users to determine their exact geographic location. These satellites can tell us exactly
where we are. Each GPS satellite has an onboard computer, an atomic clock, and a radio. Since it knows its
orbit and time, it constantly informs about its changing position and time. On the ground, every GPS receiver
has a microprocessor that “triangulates” its position by controlling it from three satellites. Once the GPS
receiver calculates its distance from three or more satellites, it can determine the user’s actual location. GPS
signals are always available. Of 31 GPS satellites in Earth orbit, 27 are in use. The minimum number of GPS
satellites required for normal operation is 24, which means that at least seven satellites are redundant. But on
Earth’s surface, it is defined that at least nine satellites are always available to send GPS signals to receivers.
Due to this redundancy, receivers can obtain stable GPS signals anywhere on Earth, and at least three or four
signals will be sufficiently strong for IO-Detection. Most of the time, in an indoor environment, the
availability of GPS satellites is not more than three. Not only that sometimes there are no satellites available
indoors due to lack of line-of-sight path to satellites. That’s why GPS does not detect indoor environments
accurately. But, if the satellite count becomes Less than three or three then it can identify the indoor
environment. Although calculating the number of GPS satellites can give glimpses about indoor and outdoor
environments but main detection result is implemented based on GPS SNR value.

Indoor
Outdoor Semi-outdoor

Figure 4.4.1 Number of visible GPS satellites that are communicated with receivers inside a building as shown in (a). Whereas the
number GPS satellite pattern is communicated with receivers outside the building as shown in (b). The number of GPS satellite
pattern that is communicated with receivers when the user moves outdoors, then return indoors as shown in (c).

In Figure 4.4.1(a), in an indoor environment, one to three signals from the GPS satellites are measured.
Similarly, in semi-outdoor and outdoor environments, this range becomes greater than 4. Typically, in Figure
4.4.1(c), this range lies between 4 and 6, which indicates a semi-outdoor environment, which is shown using
two rectangles. In this experiment, we also observed that the highest number of signals coming from GPS
satellites in outdoor Environments is 10. The total number of signals that come from GPS satellites finds
distinct patterns in indoor/outdoor environments. Figure 4.4.1 (c) indicates the decreased rate of signal that
comes from GPS satellites as the user walks from outdoor to indoor environments

17
 . (a) Indoor GPS Average SNR (b) Outdoor GPS Average SNR
25 40
20 30
15
10 20
5 10
0 0
0 10 20 30 40 0 5 10 15 20 25 30
c) Outdoor to indoor GPS SNR
30
GPS Average SNR

20

10
outdoor semioutdoor Indoor
0
0 5 10 15 20 25 30
Time(Sec)

Figure 4.4.2 Average GPS signal-to-noise-ratio pattern in an indoor environment as shown in (a). Whereas the Average GPS signal-
to-noise-ratio pattern is outside the building as shown in (b). Average GPS SNR pattern when the user moves outdoors, then returns
indoors as shown in (c).

As the term indicates, SNR is a ratio of signal power to the noise floor of GPS observation and has
conventionally been used only as a measure of receiver tracking efficacy or for comparison of signal
strengths between channels and satellites. SNR is calculated using the code tracking loop. Tracking loops are
used in a closed loop to follow continuously the code and carrier parameters of the incoming signal. The
SNR of all GPS signals is measured every 1 second, and the average SNR of GPS is measured every 10
seconds. If the SNR average of GPS is more than 20 dB and the GPS module of the terminal can
communicate with more than 6 satellites, it is judged that positioning is possible and the state remains in the
outdoors state, which is shown in figure 4.4.2 (b). In the outdoor environment, the SNR becomes much
higher, varying from 25 to 42 due to the clear line-of-sight paths between the phones and GPS satellites. In
the semi-outdoor environment, although we may sometimes observe GPS signals from more than four
satellites, typically the SNR of GPS signals is not high enough to ensure accurate localization. Figure 4.4.2
(a) shows that the average SNR of GPS indoors is less than 18 and the GPS module of the terminal can
communicate with less than 3 satellites.

18
 intdoor outdoor
40
GPS Average SNR

30
20
10
0
0 5 10 15 2 0 25 30 35
Time(seconds )
(a) Average GPS SNR measurement in different scenes. The GPS SNR readings are discrete but different for different environments.
(b) outdoor - indoor - outdoor GPS Signal SNR Numer of satellite
40
GPS signal

20

SNR not available SNR

0
1
4
7
10
13
16
19
22
25
28
31
34
37
40
43
46
49
52
55
58
61
64
67
70
73
76
79
82
85
88
91
94
97
100
103
106
Time

(b) GPS signal pattern when the user moves indoors, then returns outdoors with a smartphone in hand and screen facing up. The
average SNR and number of GPS satellites communicated with receivers are jointly plotted in this graph.

4.4.3 GPS signal measurement

The GPS module can work with more than four visible satellites. However, even with four satellite signals,
the localization accuracy varied dramatically in our experiment. With more than six visible satellites, the
localization error is around 20 meters. We also observe that more visible satellites (e.g., >9) yield less
marginal improvements in localization accuracy. With 10 GPS satellites, the localization error can be within
10 meters. In the indoor state, GPS positioning is stopped because GPS is not available. Moreover, GPS can
sometimes get a satellite fixed indoors (e.g. when the user is close to a door or window). Figure 4.4.3 (b)
plots the GPS signal pattern and its average SNR in an example scenario in which a user walks outside of the
office, passing through a corridor. In particular, the user walks from the 1st second to the 15th second in an
outdoor environment, continues walking from the 15th second to the 70th second inside the building, then
walks along the corridor from the 70th second to the 80th second, and after that, the user comes outside the
building and stops walking from 85 seconds to 100 seconds. In Figure 4.4.3(b), in the first 15 seconds, we
find that the average SNR of GPS in the outdoor environment is higher than 20 dB. Between 15 and 27
seconds, the average SNR value decreases dramatically. After 27 seconds, the average SNR becomes null
due to the unavailability of GPS satellites. Between 50 and 60 seconds, the SNR of GPS is available due to
one or more satellite communications with receivers. After 60 seconds, the SNR once again disappeared due
to no satellite being available. From 70 seconds to 85 seconds, the user passes through a corridor, which is a
semi-outdoor environment. In this situation, the average value of SNR is around 19–22 dB. After 85 seconds,
when the user starts working in outdoor environments, the SNR of GPS dramatically increases due to the
greater number of satellites communicating with receivers. We find that the average SNR is very high when
the user moves from the 1st second to the 15th second and from the 85th second to the 100th second. Hence,
indoors, when GPS is not available, other indoor positioning techniques take place. In summary, the
experiment results demonstrate that GPS availability and localization accuracy are highly correlated to the
type of environment. Yet solely reading such availability from the GPS module itself can take up to minutes
and consume much extra energy in scanning the satellites.

19
4.5. Wi-Fi AP Detector
With the ubiquity of Wi-Fi-enabled smartphones and large-scale access point arrangements, Wi-Fi-based
detecting is one of the most promising indoor/outdoor detection signatures. Due to GPS limitations in indoor
environments, the indoor localization system depends on the received signal strength (RSSI) of each Wi-Fi
access point. Wi-Fi received signal strength indicator (RSSI) upgrades are one of the most attractive
techniques due to their dependence on the ubiquitously deployed infrastructure. The accuracy of the Wi-Fi
location relies on the number of access points. This module employments the complete Wi-Fi infrastructure,
which incorporates not only Wi-Fi routers but also versatile phones and tablets. By utilizing Wi-Fi access
points, this system can effortlessly find all Wi-Fi-powered devices' locations and monitor their movements
inside the building.
This Wi-Fi AP Detector does a Wi-Fi scan every second and lists all the access points detected. Every list
row shows the access point’s SSID, BSSID, current RSSI, and frequency. The number of Wi-Fi APs makes a
significant difference between indoor and outdoor environments, as shown in Figure 4.5.1. During
experiments, we implemented more than 250 scans through different sites in indoor areas like Universities,
and homes and in the surrounding outdoor areas with different mobility scenarios (standing and walking).
The number of all visible detected Wi-Fi nodes is calculated based on their MAC addresses (BSSID), where
each AP has a unique MAC address. Figure 4.5.1 says clearly that, during the scan, outdoor sites exhibit all
visible Wi-Fi AP density confined between thresholds α1 and α2, and sometimes higher than threshold α1
reaching 39 AP. In contrast, indoor sites often exhibit a visible Wi-Fi AP density lower than threshold α2,
sometimes less than 5 APs.

35
Wi-Fi APs Number

30
25

20 α1
15
10 α2
5
-50 50 150 250 350 450 550
Scan Index

Figure 4.5.1: Distribution of all visible Wi-Fi access points number density at surrounding urban/indoor environments

0 WIFI_Max_RSSI
0 100 200 300 400 500 600
Wi-Fi Max RSSI

-30
indoor
-60
outdoor
outdoor
-90

-120
Scan Index

Figure 4.5.2: Distribution of Maximum RSSI for all visible Wi-Fi access points in surrounding outdoor and indoor environments

The Wi-Fi detector module collects all the RSSI obtained and finds the maximum and average values of
RSSI. APs exhibit higher RSSI in indoor areas than in outdoor areas. This is because users inside the
building will be nearby and in line of sight with many of the indoor nodes, where the correctness percentage
varies according
20
to the user mobility scenario. On the other hand, APs exhibit very weak RSSI records but are still heard by
clients’ mobiles when passing by outside areas like buildings and narrow streets. Figure 4.5.2 shows that in
an indoor environment, the maximum RSSI value is higher than -67 dB, which indicates a strong Wi-Fi
signal. As shown on a chart, the X-axis represents time, and the Y-axis represents the signal level in dB. But
using this maximum RSSI, we cannot reach any decision because sometimes, when the Wi-Fi router is close
to the outdoor scene, the maximum RSSI range falls under -55 dB to -69 dB. The RSSI higher than -50
indicates we are very close to or in the same room as the Wi-Fi router. The possibility of which is very low.
But combining the maximum RSSI, the average RSSI, and the number of detected Wi-Fi APs (less than -67
dB) gives us a clear indication of the indoor environment. Which is shown in the IO detection algorithm.
From Figure 4.5.4, we can understand the difference between extremely strong, strong, and weak signals.
RSSI range Observation
RSSI<-90 dBm Extremely weak Approaching or drowning in the background
signal transmissions and causing serious interference
with the signal.
RSSI<-80 dBm Not a good signal Minimum signal strength for basic
connectivity
RSSI<-70 dBm Okay Minimum signal strength required for decent
packet delivery
-67dBm <RSSI<-55 dBm fairy strong signal Minimum signal strength for most business
applications.
-55 dBm <RSSI<-30 dBm Very strong signal The client would see the AP or vice versa.
RSSI>-30dBm Excellent Maximum achievable signal strength.
Table 1: Wi-Fi RSSI range
-60
RSSI Average of Wi-

-65 0 70 140 210 280

-70
-75
Fi APs

-80
-85
β1
-90 β2
-95
-100
Scan Index
Figure 4.5.4: Distribution of the RSSI average for all visible Wi-Fi access points in surrounding outdoor and indoor environments

We can measure, in every scan, the received signal strength from all visible Wi-Fi access points. Then we
aimed, in addition to estimating their average and number, to infer other useful features that classify Wi-Fi
APs seen by smartphones. For each access point, if its RSSI value falls within the strong RSSI range, it is
called the 1st AP class. Figure 4.5.3 shows that the average of all visible Wi-Fi APs RSSI exhibits distinct
signatures for indoor and outdoor environments. We observe that the most commonly collected average
values of all visible Wi-Fi APs RSSI in indoor sites are higher than the 1 threshold. This is seen as logical
because the majority of these nodes should provide strong RSSI readings to meet connectivity
considerations. Unlike average values collected at outdoor sites, most of them are lower than 2 because most
detected signals are weak signals that can still be heard due to buildings' structure and distance
considerations. During tracing all tested indoor and outdoor sites, we observed that almost all indoor sites
exhibit a APs number of the 1st order class (which has an RSSI higher than -67 dB) that is larger than at
outdoor sites. This occurs because the probability of the expectation that the user, during indoor movements,
will become very close to such APs is high.

21
4.6 Overall estimation
This proposed method uses the estimation by illumination, the estimation by magnetism, the estimation by
GPS, and the estimation by Wi-Fi AP concurrently and acquires each estimation result's reliability. Each of
the four subdetectors has unique advantages and disadvantages. For example, a light detector requires the
mobile phone to be exposed to space. If mobile phones are inside a bag or pocket, a light detector cannot
provide accurate detection results. The Wi-Fi detector needs sufficient Wi-Fi AP signals to detect the
ambient context. The magnetic detector is only available when the user is moving around, such that the
magnetic disturbance inside the building can be exploited. The GPS detector sometimes stops working in an
indoor environment. We refer to the four individual detectors as subdetectors and integrate them to output an
arbitrated decision. As described in the IO-Detector algorithm, each subdetector outputs the possible
environment types and associated confidence levels for them. For each possible environment type, we sum
the confidence levels from the four subdetectors. If the four estimation results are different, the proposed
method considers the result with the highest reliability to be the estimation result. For example, in the
outdoor state, the method of estimation always relies on GPS because the reliability of GPS in the outdoors is
higher than other estimation methods. Similarly, indoors, the estimation method depends on Wi-Fi.
Whenever Wi-Fi is unavailable, the estimation method relies on either light intensity or magnetism variance.
For example, in the case of rural outdoor scenarios, a light intensity module will be more useful than Wi-Fi
because Wi-Fi RSSI will be nonexistent and more difficult to hear than those in urban or indoor
environments.
Finally, based on those observations and challenges, we construct the IO-detection algorithm as shown in
algorithm 4.6.

22
 Algorithm 4.6 IO-Detection algorithm
>> Li: Current light intensity measurement.
>> Mg: Current Magnetism intensity measurement.
>>Mag_Variance: Calculate magnetism variance every 10 seconds.
>>GPS_SNR: GPS signal-to-noise-ratio per sec.
>> GPS_Count: All visible GPS satellites per scan.
>> Avg_SNR: SNR Average of each GPS signal per sec (dB).
>> Avg_ Trend: Calculate Average SNR Trend per 10 sec (dB).
>> WAP_D: All visible Wi-Fi APs density per scan.
>> Max_RSSI: RSSI max of All visible Wi-Fi APs per scan (dB)
>> WAP_Avg: RSSI Average of All visible Wi-Fi APs per scan (dB)
>> 1st_APs: Number of Wi-Fi AP whose RSSI>-67 per scan.

1. Check Proximity sensor availability on Smartphone

2. if available, Get Li, Get Mg, and Check the Current Time System clock
3. if (Li >=2000) and Current Time is a Daytime
4. OUTDOOR CONFIDENCE =+6
5. Else if (Li ≤ 1000) and Current Time is a Daytime
6. INDOOR CONFIDENCE =+6
7. if(GPS_SNR>=23 & GPS_Count>=4 OR GPS_SNR>18 & GPS_Trend<6 )
8. OUTDOOR CONFIDENCE =+8
9. if (GPS_SNR<18 && GPS_Count<3)
10. INDOOR CONFIDENCE =+8
11. if ((WAP_D ≥8 & 1st_Aps>3) or (Max_RSSI≥ -50 & WAP_Avg ≥ -67))
12. INDOOR CONFIDENCE =+9
13. else OUTDOOR_ CONFIDENCE =+6
14.if (Mag_Variance<18)
15. OUTDOOR_ CONFIDENCE =+3
16. else INDOOR CONFIDENCE=+3
17.if(INDOOR_ CONFIDENCE >OUTDOOR_ CONFIDENCE)
18. Detected Environment is Indoor
19. else Detected Environment is Indoor
20. End.

23
Chapter 5

Experiment result
5.1 Experimental Setup
Throughout our experiments, measurements are collected using different mobile users walking multiple trips
over different paths to assess the performance of each IO-Detector module over diverse scenarios. Some trips
are made indoors (inside buildings, e.g., homes, universities, and malls) and others are made in outdoor areas
(e.g., open areas). We implemented and evaluated the IO-detector prototype on the Android Platform and
tested its performance on a smartphone as an Android application (Realme-8). This phone is equipped with
all the sensors required for IO-Detection such as proximity, time, light sensor, accelerometer, Magnetometer,
GPS, and Wi-Fi.
Outdoor Semi-outdoor Indoor

Fig. 5.1 IO-Detector application

5.2 Performance
For the evaluation of the estimation accuracy, it is necessary to record the correct location of the terminal
(correct answer data). At every movement between outdoor and indoors, I select the correct answer data by
selecting the current position button and storing it. Therefore we made the process to record indoor or
outdoor, and at every movement between indoor and outdoor, the user selects the current position using the
current position button and records whether it is indoor or outdoor. This record becomes the correct answer
data. We compare the estimated result with this correct answer data and calculate the precision ratio. In this
section, we show the detection performance of the three individual subdetectors as well as the aggregated IO-
Detector.

24
 5.2.1. Performance of Subdetectors.

indoor indoor indoor indoor

Semi-outdoor/outdoor Semi-outdoor/outdoor Semi-outdoor/outdoor Semi-outdoor/outdoor
100 80 120 100

80 100 80
60
Detection ratio(%)

80
60 60
40 60
40 40
40
20 20
20 20
0 0 0 0
IN SEMI/OUT IN SEMI/OUT IN SEMI/OUT
IN SEMI/OUT
(a)Light Detector (b)Magtenism (c)GPS Detector (d)WiFi Detector
Detector

Fig. 5.2. Detection performance of 4 subdetectors. For each subdetector, we report the performance under 3 kinds of scenarios. For
example, we test the light detector in indoor environments and outdoor/ semi-outdoor environments, respectively.

To evaluate the contribution of each detector (i.e., light detector, GPS detector, Wi-Fi detector, and
magnetism detector), we examine the detection performance independently in Figure 5.2 using correct
answer data. Each detector reports the environment type with the highest confidence level after the local
computation. The light detector is available when there are clear paths between mobile phones and ambient
light sources. Figure 5.2(a) shows the detection performance of the light detector. We find that the light
detector can effectively distinguish the indoor environment from the outdoor environment. In Figure 5.2(a),
in the indoor environment, the detection accuracy is around 90%. When the phones are in the outdoor
environment, the detection accuracy is around 92%. However, it was not able to be estimated with accuracy
at the time without a clear illumination difference between indoor and outdoor such as evening. This is
because a false estimate of the indoors occurs by the very small difference of the illumination when the
subject stopped in the shade for a long time. Figure 5.2(b) shows the detection performance of GPS detectors
that can significantly differentiate between indoor and outdoor environments. The GPS detector provides
high accuracy for both indoor and outdoor environments with average power consumption. GPS detector
provides a detection accuracy exceeding 90%. Figure 5.2(c) shows the detection performance of the Wi-Fi
detector that classifies the indoor environment from the outdoor environment. Figure 5.1 shows the result of
applying the Wi-Fi-based module algorithm on measured Wi-Fi measurements (RSSI, BSSID, and SSID).
Our experiments mainly cover the campus. In such experiment settings, the Wi–Fi–based detection performs
with 85% accuracy. An average error ratio is around 7% in both indoor and outdoor experiments. This
average error detection significantly decreases when the Wi- Fi APs are available and more ubiquitous in the
ambient environment. We obtained quite a close performance of the Wi-Fi detector compared with that of
the light detector. We note that both the light detector and Wi-Fi detector can effectively classify the indoor
environment from the outdoor environment. On the other hand, the magnetism detector can enhance the
detection capability of the IO-Detector in classifying the outdoor environment. Figure 5.2(d) plots the
performance of the magnetism detector. The estimation only by the magnetism was able to attain around
60% of the precision ratio on average. Because it was estimated to be the outdoor by mistake indoor with the
magnetic variation, the precision ratio did not rise.

25
 Table 3: Scenario of the evaluation experiments
Ex. Date Time Place Environment Weather
No.
Ex.1 24/7/23 17.37-17.40 Central library to impact outdoor sunny
center Jadavpur university
Ex.2 24/7/23 16.31-16.35 Within the campus indoor sunny
Ex.3 23/7/23 17.48-17.54 home Indoor Cloudy/rainy
Ex.4 23/7/23 18.17-18.22 Outside the Residental house outdoor Cloudy/rainy
Ex.5 25/7/23 10.04-10.10 Within the campus Semi-outdoor sunny

Table 4: Result of the evaluation experiments

Detected Ex. No. No of No. of Error Proposed
Environment detected environment method
Environment Detection accuracy

INDOOR Ex.2 70 3 95.71%

Ex.3 283 26 90.81%
Ex.1 83 6 92.77%
Outdoor Ex.4 240 5 97.91%

Semi-outdoor Ex.5 40 16 60%

DETECTION RESULT
indoor semioutdoor outdoor

93% 95%
60%

INDOOR SEMIOUTDOOR OUTDOOR

Fig. 5.3. Detection precision
4.2.2. Performance of Aggregated IO-Detector.

The evaluation experiment is based on 5 scenarios in Table 3. In these scenarios, the subject walks holding
the terminal in his hand. Table 4 shows the results of the experiments. The data collected process of Ex.1
was conducted in an outdoor environment as shown in Table 3. Similarly, other experiments were conducted
in their designated environment as maintained in Table 3. Of the 83 detected environment samples in Ex.1,
six detected samples indicated error environments. From this, we calculated the proposed method accuracy
of Ex.1 which is 92.77%. Similarly, we have calculated the detection accuracy for our proposed methods for
the rest of the experiments. From Table 4, we get the detection accuracy of each environment for each
conducted experiment. Since a semi-indoor environment is not available everywhere that’s why we exclude
it from overall detection accuracy. By adding these four indoor/outdoor detection accuracy we get the overall
detection accuracy. The overall detection accuracy of the IO-Detector is about 95% which is higher than
other independent estimations. When the four subdetectors are aggregated as the IO-Detector, there is
improvement in detection accuracy for all different types of indoor/outdoor environments, but not much.
Because the proposed Aggregated IO-Detector method can use the other estimate technique in a situation in
which one estimated technique is weak, the method can keep estimation with high precision. For
example, it is the

26
situation such as at the time of the outdoor movement with no Wi-Fi AP found in the estimation by the Wi-
Fi- detector and the smartphone is in the subject’s pocket which indicates the estimation by magnetism is
difficult or the situation in the evening in the estimation by the illumination. In this case, Estimation by GPS
Detector provides high accuracy. Compared with less than 60% detection accuracy of individual detectors, in
the aggregated IO-Detector both the precision and the recall are consistently above 90% for the indoor and
outdoor environment type. The experiment results suggest that the IO-Detector accurately classifies the
indoor/outdoor environments for most cases.

27
Chapter 6

Conclusion and Future Work

In this paper, we proposed the indoor outdoor detection system using a smartphone maintaining the
detection precision. By intelligently aggregating the subdetectors, IO-Detector achieves prompt and accurate
detection results in various times and environments. In future work, we need a comparative evaluation of the
precision and power consumption.

28
References
[1] S. S. A. E. M. F. M. Y. a. R. R. C. H. Wang, "UnLoc: No Need to War-Drive:Unsupervised Indoor
Localization," in In Proc. MobiSys, 2012.

[2] J. L. H. H. S. S. a. H. L. S. Jeon, "Indoor WPS/PDR Performance Enhancement Using Map Matching

Algorithm with Mobile Phone," in IEEE/ION Position, Location, and Navigation Symposium PLANS,
2014.

[3] Y. H. W. B. G. X. S. L. T. X. K. D. a. Q. W. J. L. Liu, A Novel Indoor Positioning Method Based on Location

Fingerprinting, International Conference on Communications, Circuits, and Systems (ICCCAS), (2013).

[4] Z. P. Z. Y. L. Z. &. S. G. (. Li M., "Iodetector: A generic service for indoor/outdoor detection".

[5] e. a. P. Zhou, “IODetector: A Generic Service for Indoor Outdoor Detection, Proc. of SenSys’12, 2012.

[6] Y. K. J. Y. a. E. S. K. S.-E. Kim, "Indoor positioning system using geomagnetic anomalies for
smartphones," in International Conference on Indoor Positioning and Indoor Navigation (IPIN), 2012.

[7] A. P. a. T. H. Fehmi Ben Abdesslem, "Less is more: Energy-efficient mobile sensing with senseless.," in
MobiHeld ’09, New York, NY, USA, 2009.ACM.

[8] J. S. H. P. a. C. H. H. Cho, Proc. 11th Int. Symp. Location-based Services (2014).

[9] R. B. T. R. Hansen, "Seamless indoor/outdoor positioniing handovver for location-based services in

streamspin," pp. 267-272, 2009.

[10] C. N. H. B. a. S. M. L. Ravindranath, Improving Wireless Network Performance Using Sensor Hints, In

Proc. USENIX NSDI, 2011.

[11] A. T. a. G. S. H. Vathsangam, "A Data-Driven Movement Model for Single Cellphone Based Indoor
Positioning," in International Conference on Body Sensor Networks (BSN), 2011.

[12] R. W. F. L. a. Y. H. Z. Xu, "AENS: Accurate and Efficient Mobile Phone Indoor Navigation System
without WiFi," in IEEE 11th International Conference on Dependable, Autonomic and Secure
Computing (DASC), 2013.

[13] P. E. L.-d. T. a. A. R. O. Canovas, in IEEE Sens. J. 17 (2017).

[14] S. P., "Radiometry and photometry in astronomy.," (2009).

[15] E. 1.-1. C., Light and lighting - lighting of workplaces - part 1: Indoor workplaces, (2002).
[16] G. J. v. B. N. L. C. H. S. &. K. V. Klakegg S., "Towards commoditized near infrared spectroscopy," in

29
 Conference on Designing Interactive Systems, 2017.
[17] P. N. a. S. T. Samuli Hemminki, "Accelerometer-based transportation mode detection on
smartphones," in In Proceedings of the 11th ACM Conference on Embedded Networked Sensor
Systems, SenSys ’13, New York, NY, USA, 2013. ACM.

[18] F. D. Y. J. G. J. &. K. V. Jiang W., "Pulse: low bitrate wireless magnetic communication for
smartphones," in Proceedings of the 2014 ACM International Joint Conference on Pervasive and
Ubiquitous Computing, ACM, (2014).

[19] U. M. Q. K. S. Mamta Bachani, Performance analysis of Proximity and Light Sensors for Smart Parking,
Kowloon Tong, Hong Kong: Elsevier B.V, 2016.

[20] M. A. a. M. Youssef, "Uptime: Ubiquitous pedestrian tracking using mobile phones," in In Wireless
Communications and Networking Conference (WCNC), 2012 IEEE, pages 3204–3209, April 2012.

[21] P. K., S., M. K. M. Valentin Radu, "Poster: Am I Indoor or Outdoor?," in MobiCom’14, Maui, Hawaii,
USA., September 7-11, 2014.

[22] M. A. M. Al-Awami, "SENSEI/O: REALISTIC UBIQUITOUS INDOOR OUTDOOR DETECTION SYSTEM

USING SMARTPHONES," 2014.

[23] J. K. a. R. G. Jeongyeup Paek, "Energy-efficient rate-adaptive gps-based positioning for smartphones,"

in MobiSys ’10, New York, NY, USA, 2010.ACM.

30