Attendance System Using a Mobile Device: Face Recognition, GPS or Both?

Abstract

— In every higher education setting in Malaysia, there are concerns about student attendance, as the current process of manual attendance taking is not only time consuming but is also inaccurate. Inconsistent attendance in class may significantly affect students’ overall academic performance. Thus, having a consistent attendance system is important. This paper proposes a mobile attendance system equipped with face recognition and a GPS locator. The face recognition adopts the Local Binary Pattern Histogram (LBPH) algorithm and retrieves the student’s location using GPS services. This project has a high potential to replace the current attendance system, as it is designed for speed and accuracy and is more convenient than the
current approach.

Full text

International Journal of Advances in Electronics and Computer Science, ISSN: 2393-2835 Volume-3, Issue-8, Aug.-2016
Attendance System Using a Mobile Device: Face Recognition, GPS or Both?
26
ATTENDANCE SYSTEM USING A MOBILE DEVICE: FACE
RECOGNITION, GPS OR BOTH?
1GEETHA BASKARAN, 2AHMAD FARHAN AZNAN
1,2The University of Nottingham (Malaysia Campus), Computer Science Department
E-mail: 1Geetha.Baskaran@nottingham.edu.my, 2farhan5157@gmail.com
Abstract— In every higher education setting in Malaysia, there are concerns about student attendance, as the current process
of manual attendance taking is not only time consuming but is also inaccurate. Inconsistent attendance in class may
significantlyaffectstudents’overall academic performance. Thus, having a consistent attendance system is important. This
paper proposesa mobile attendance system equipped with face recognition and a GPS locator. The face recognition adopts the
Local Binary Pattern Histogram (LBPH) algorithm and retrieves thestudent’s location using GPS services. This project has a
high potential to replace the current attendance system, as it is designed for speed and accuracy and is moreconvenient than the
current approach.
Keywords— Mobile attendance, Face recognition, GPS, Local Binary Pattern.
I. INTRODUCTION
As a student, it is necessary to attend regularly all
lectures, tutorials and lab sessions listed in the
timetable. Doing so enables student to learn
effectively across the semester with the designed
syllabus. Some might argue that independent learning
is the best way for students to learn and that students
have the right to manage their own time, even if this
means missing class. However, considering the
amount of money students are paying for their
education, and the fact that lax attendance systems are
known toaffect particular students’ studies and the
university’s reputation [1],attendance recordsare
important to understand student progress and
development. In some institutions, without a certain
percentage of attendance, studentsare not allowed to
sit for an examination, while in some other
institutions,attendance is part of the continuous
assessment.
The classic attendance system of calling students’
names and recording their presence on paper is easily
manipulated by students [2],who know how to abuse
the system and have their friends record their
attendance falsely. The system is also time consuming
and may adversely affect students’ learning
experience [3]. Imagine how long it would take to
register attendance in a class of 100–300 students
using this method. Further, it would require
preparation on the part of the attendance taker, which
is tedious, and the lecturer would need to monitor
students manuallyto detect cases of dishonesty. To
solve this problem, some universitieshave introduced
the Moodle attendance system,where an attendance
link is provided for a short period for students to
update their attendance. This method requires students
to login to a device using their own account. However,
while this method improves efficacy, some
problemsremain. For example, students can mark their
attendance from outside the class or university with
the right timing. Therefore, it is essential to find a new
approach to overcome the problems of the current
methods and to make it more convenient for
lecturersto record their students’ attendance.
Previous attempts at automated time and attendance
systems have used electronic tags, barcoded badges,
magnetic strip cards, biometrics (vein reader, hand
geometry, fingerprint or facial) and touch screens [4]
in place of paper cards. Such attendance systems try to
overcome the aforementioned problems by ensuring
studentsaredirectly interacting with the device and that
the device is in the particular class in which the student
should be. Conversely, the aim of this project is to
create an attendance system that allows students to
record their attendance using their own mobile device,
with the help of face recognition technology and a
GPS locator.Our proposed attendance system does not
require any kind of peripheral device other than
students’ own smartphones,thereby reducing
computational time and avoiding the cost of placing
physical devices in classes.
II. PREVIOUS WORK
Several techniques and methods have been accepted to
effectively monitor students’ attendance. Shoewuet al.
[5] proposed a cost-effective computer-based
embedded attendance management system that
allowed the electric monitoring of attendance using
anelectronic card. These cards, which contain all
necessary information on the individual, are inserted
into a machine that records the time and other
information to a server. In another example, Cheng et
al. [6] designed and implemented a system that applies
user identification and a passwordfor authentication.
However, the issue with these electronic card or
password-based systems is that they allow for the
sharing or dishonest use of the cards or passwords.
This problem can be addressed by using a biometric
recognition system,such asfingerprint or iris
recognition.A system was proposed and implemented
by the authors in [7] and [8] for using fingerprint scans

International Journal of Advances in Electronics and Computer Science, ISSN: 2393-2835 Volume-3, Issue-8, Aug.-2016
Attendance System Using a Mobile Device: Face Recognition, GPS or Both?
27
to record attendance and generate reports after a fixed
duration. To have their attendance verified,
individuals simply had toinsert theirfingers into a
fingerprint reader.
In another attempt to address the problem of misuse of
electronic attendance-taking systems, Kadry et al [9]
proposed a wireless attendance management system
using an individual’s iris, which is unique, for
authentication. In this system,a scanner is used to scan
the iris and automatically log in the person.
Unlikefingerprints, the iris is more preserved from the
external environment. However, both fingerprint- and
iris-recognition-based approachesrequire extra
devices and scanners,usually connected to a server.
In radio frequency identification (RFID)-based
methods, attendance is recorded in the same way as for
the fingerprint reader, with the only difference being
the tools used;that is, the RFID card [10]. The RFID
card stores user’s information on the card as data. This
data is encrypted into the card, which is then used as a
key to record when the user arrives [11].
In our work, we address the problem of the misuse of
electronic attendance-taking systems byusing the
internet connectivity of smartphones to monitorthe
presence or attendance of an individual.
Smartphone-based monitoring systemsprevent the
expense of additional scanning devicesby leveraging
on the fact that almost all studentsown asmartphone.
In our system, an area is fixed for every student. When
he or she enters or exitsthat area, a time stamp is saved
and the system calculates the duration of any particular
student residingwithin the area.
III. AGILE METHODOLOGY
A. Extreme Programming (XP)
Extreme programming (XP) is one of the agile
methods that emphasize iterative developmentand has
been very effectivein producing high-quality software
in real-worldprojects with strict time constraints.Table
1 maps the XP practices, distinguishing the XP
practices which address the software quality andthose
which address the development process quality. We
used this mapping to address the software quality,
emphasize technical and code-oriented aspects,
whereas the XP practices,which address the
development process, emphasize human and social
aspects. Needless to say that bothaspects are important
and there is a synergetic relationship among them [12].
Table 1: XP practices mapping with respect to
quality subjects
B. Testing
We piloted our program first in University of
Nottingham Malaysia Campus (UNMC) and
Polytechnic Sultan Abdul Halim Mu'adzam Shah
located in Jitra, Kedah. We have tested not just in the
smallest room in UNMC as well as the largest room in
UNMC. We conducted the testing usingfourdevices
:Samsung S advance, Oneplus one, Galaxy Mega 2
and Samsung S5. The results are identical. The devices
are all similar inits process time which is less than 1
second.
IV. PROPOSED SOLUTION APPROACH
A. How the Program Works
The program needs to be installedon an android device
with an active internet connection, GPS and a camera.
Users need to follow some simple steps to enable the
program to update their attendance record,including
permitting the program to obtain their current location
and providing the necessary input to allow the
program to recognise the student’s face. This program
can only recognise one face at a time.
The primary objective of the program is to be able to
take attendance using students’ mobile devices
without inaccuracy. For the project to succeed,it must
employ a location tracker and face recognition to
deliver a reliable attendance system. The face
recognition requires the student to have direct
interaction with the device, while the GPS locator
specifies the device’s location. Students also benefit
from this system,as they can check their attendance
status for their current classes. Since student can use
their own mobile device, which they can bring
anywhere,the system offers excellent mobility.
To summarise, this program requires the following
core functions:
 Face recognition
 A classroom locator
 The ability to register attendance and allow
students to access their attendance record.
1) FacialRecognition Technology
A camera is needed to use facial recognition. Before
deploying the application to users, it must be
initialised with the required dataset images (facial
images of the students),which will be processed at the
start of the program. To ensure the optimal speed of
the program, the dataset images must be minimised.
The best way of doing this is by separating the dataset
images by course. Thus, every student in each course
would use the same application, except it will have
been initialised with different images. Once the
program has been loaded with the student images, it
will be able to recognise students’ faces by using an
appropriate algorithmto compare the current frame
image with the one that has been initialised.
Initialising the images in the program before
generating the installer provides much greater
reliability because students cannot easily alter the
initialised images.

International Journal of Advances in Electronics and Computer Science, ISSN: 2393-2835 Volume-3, Issue-8, Aug.-2016
Attendance System Using a Mobile Device: Face Recognition, GPS or Both?
28
2) Classroom or Location Locator
GPS coordinatesare requiredfor the program to
instantly determine the student’s current location,
based on the coordinates received. Using GPS, we can
obtain both x- and y-coordinates up to 6 decimal
points with the help of ground and space satellites. The
location becomes even more accurate with the help of
cellular network providers. To interpret the
coordinates, the program must be integrated with
Google Maps APIs so that users can view the visual
location of the coordinates receive. A credential is
required to use the Google Maps APIs service,which
can be obtainedby placing a request through Google
console.
3) Register Attendance
The system automatically updatesattendance in the
database for any faces that the program could
recognise. Students’ mobile devices are remotely
connected to the local database. The information
updatedis student name, x-coordinate, y-coordinate,
classroom and timestamp. In addition to registering
attendance, the application allows students to access
their attendance record. The x- and y-coordinates
stored during attendance registration can also be
retrieved to show the location on Google Maps.The
workflow for this program is outlined in Section B
(Fig.1).
C. Overall Workflow of Program
Fig.1. Attendance system overall workflow
D. Local Binary Pattern Histogram
Only a few facerecognition algorithms are provided in
the OpenCV library,including Local Binary Pattern
Histograms (LBPH), Eigenfaces and Fisherfaces. This
project uses LBPH, which takes a different approach
comparedto the other methods. In LBPH,
characterisation of featuresis done locally,whereas
other approaches process the image as a whole.
TheLBP algorithm comes from a visual descriptor for
pattern classification mainly used in computer vision.
In this project, training imageswere set to 128 X 128
pixels. It is important to maintain the image size to
avoid affecting the face recognition rate. This is
because the LBPH algorithm is highly prone to scaled
images. That is, once the algorithm extracts a feature,
the program can only identify the person when
provided with an image at the same scale (in pixels).
The first requirement of the LBPH algorithm or the
pre-processing procedure is to convert the image to
grayscale mode. Grayscale imagesare not images in
black and white or binary images; instead, grayscale
mode is a series of numbers, each of which represents
a different intensity level. Having images in grayscale
mode represents a significant advantage when using
the LBPH algorithm,as the image can be treated as a
vector to extract valuable information. Next, for each
pixel in the grayscale image, we select a
neighbouringpixel of size 8 surrounding the centre
pixel. The LBP value is calculated based on the centre
value by thresholding it to a 3 X 3 array. The intensity
level threshold is set to 8. A more formal description
of the LBP operator can be given as equation
(1),where the notation (P, R) denotes a neighbourhood
of P sampling points on a circle of radius R:
LBP,(,)=S(


−)2(1)
Formally, given a pixel at (Xc,Yc), the resulting LBP
can be expressed in decimal form as in equation
(1),where intensity ip and ic are respectively gray-level
values of the central pixel and P surrounding pixels in
the circle neighbourhood with a radius R, and function
s(x) is defined as:
The operator LBP (P, R) creates 2p different output
values, matching to 2p different binary patterns
formed by P pixels in the neighborhood.The basic
LBP operator is invariant to monotonic gray-scale
transformations maintaining pixel strength order in the
local neighborhoods. The histogram of LBP labels
calculated over a region can be exploited as a texture
descriptor.Fig.2 is an example of calculating the LBP
value with a neighbouring size 8 pixel.

International Journal of Advances in Electronics and Computer Science, ISSN: 2393-2835 Volume-3, Issue-8, Aug.-2016
Attendance System Using a Mobile Device: Face Recognition, GPS or Both?
29
Input Output
Fig.2.Calculation of LBP value—size 8 pixel
In the above figure, the value 4 is the centre pixel.
Note that each of the values represents acolour
intensity. Theexpected output is an 8 binary digit for
each pixel LBP calculation. After performing the LBP
calculation, the value is stored in a 2D array with the
exact same dimension as the input image. With 8
adjacent pixels converted to binary digits, we have a
total of 2^8 = 256 possible combinations of local
binary patterns. The stored result in an 8-bit array can
be processed to obtain a decimal value. This process is
visualised in Fig.3.
Fig.3.LBP value stored in a 2D array
For the purpose of illustration, we start at the top right
and move clockwise (the blue boxes indicate the
sequence) to accumulate the binary string. Note that
the sequence of collecting the binary string does not
matter provided we use the same sequence for all other
Local Binary Pattern (LBP) calculations.
Fig.4.The output value from the original image
Fig. 4 illustrates on the right the output value from the
original image on the left. The process of thresholding,
accumulating binary strings and storing the calculated
LBP value is repeated for each pixel in the input
image. After obtaining all LBP values for each pixel,
the next step is the histogram. The histogram
represents the number of times each LBP pattern
occurs that acts as a feature vector. TheLBPH
algorithm represents the local structure of an image by
calculating the histogram efficiently and summarising
the histogram across different blocks.
What makes the LBPH algorithm different from others
is that each image in the dataset is locally
characterised. When a new image is provided, the
same feature analysis is performed and the result is
compared to the dataset images. The purpose is to
analyse the local structure in an image by comparing
each pixel with each adjacent pixel. For example,a
pixel is taken as centre and then thresholded against its
neighbours. A neighbour with greater or equal
intensity is denotedby1,with lesser intensity denoted
as 0. This process resultsin a binary number foreach
pixel (e.g., 1101011). Theoretically, with 8
surrounding pixels, there are2^8 possible
combinationsofLBPs.
To summarise, the steps to createLBP histogramsare
as follows:
1. Convert image to grayscale
2. Calculate the LBP for each pixel
3. Create a histogram based on each of the LBP
values
4. When new faces areprovided, generate the
LBP histogram exactly as was done forthe
trained image.
5. Recognition comes when a new histogram
matches the histogram pattern of a trained
image.
E. GPS
Mobile phones equipped with a GPS receiver
arereadilyavailable onthe market. The General Packet
Radio Service (GPRS) is currently one of the best and
cheapest communication modes available.The
attendance system is deployed onthis kind of mobile
phone, which is supported to perform all
requiredoperations. When the application is started on
a user’smobile phone for the first time, they are
prompted to register.Thereafter, the user opensthe
software by entering theirusername and password.
Whenthe user enters their username and
password,thesearechecked for authenticity. If not
authenticated, the user is prompted with a message of
wrong username and password and may re-enter their
log in details.
V. DEVELOPMENT ENVIRONMENT
The Java-enforced object-oriented programming
language developed by Sun Microsystems was chosen
for implementing this mobile program. It is designed

International Journal of Advances in Electronics and Computer Science, ISSN: 2393-2835 Volume-3, Issue-8, Aug.-2016
Attendance System Using a Mobile Device: Face Recognition, GPS or Both?
30
to have as few implementation dependencies as
possible. We have used Java version 1.7.0, designed
for 64-bit operating systems. The reason for choosing
Java is the platform’s independence criteria. A Java
program can be developed and run on almost any
computer that comes with the Java Runtime
Environment. Java applications are usually converted
into a special byte code so that they can run on any
virtual Java machine, regardless of the computer
architecture. Moreover, the PHP language is a
hypertext processor widely used by web developers,
particularly in server-side scripting language. In this
project, we implement a short PHP that acts as a
translator between the database and the application
itself. The PHP script will be placed in the XAMPP
local web server.
The developed software required a database to store
thedata and managecommunication between the
application and database. XAMPP wasthe perfect
solution. XML stands for Extensible Markup
Language, which functions as a set of rules for
encoding format. XML is also used in android
development for purposes such as defining
applications’ user interfaces, describing components
of the system and for minor purposessuch as replacing
hard-coded strings with a single string.
Forthe development environment,we chose the newest
modern integrated development environment, Android
Studio,developed by Jet Brains. Android Studio was
designed specifically for android development
purposes and has taken over end support for Eclipse
(another integrated development environment for
android)from Google. The official language of
Android Studio is Java,as alarge part of android
iswritten in Java and its API isintended to be called
from Java. Android Studiois a very useful tool because
each of itsmodules is independent andcan be run,
tested and debugged without affecting another
module. Moreover, Android Studio provides
improved features for interface design, includinga
drag and drop feature and a delivering mechanism for
interaction with resources and multi-tasking. Android
Studio alsohelpsdevelopersby adding an external
library and providing complete support forJunit and
android testing. Forthis project, Android Studio
version 1.5 was used throughoutthe software
development process.
IV. RESULTS AND DISCUSSION
The system proposed is a real-time system. The
experimental results showed that the acceptance
detection ratio of our suggested algorithm ranged from
75% to 95%. The result of the analysis process is
presented below:
F. Main Menu and Login
Fig.5. Main Menu
Fig.5 shows the main menu of our system. We design
the interface design in a simplicity way by mainly
focusing on user-friendly aspect. The sequence
numbered presented in fig.5illustrate step of taking
attendance.However, student can navigate to any other
function available such as locating current location,
taking attendance with face recognition and accessing
attendance log as they want. The enrolment and
registration phase is an administrative phase in which
the administrator (staff) needs to log in as shown in
fig.6. The studentsface photos as well as the other
bio-data are stored for the first time into the database
for student registration. The student can login to the
main menu through the student login.
Fig.6. Thesystem login
G. Test Data
Fig.7. Faces in different lighting and distance

International Journal of Advances in Electronics and Computer Science, ISSN: 2393-2835 Volume-3, Issue-8, Aug.-2016
Attendance System Using a Mobile Device: Face Recognition, GPS or Both?
31
Fig.7 shows some sample faces captured during the
registration phase. There are four different samples of
lighting, angle and distance taken.
H. Location detector
Fig.8. Left figure: Current Location, Right figure: Building
Markers
The next interface is the location map. Once the
student click “Your Location” button, it will directly
open google map interface. Refer to fig.8 (left), the
maps is programmed to zoom at user’s current
location. However, if the student wish to control the
range of the map displayed, they able to that with the
zoom in and out button. To increase the student
understanding of the map, the map has been flagged
for each building such as in fig.8 (right).
I. Face Recognition
Fig.9. Face Recognition Interface
Next interface is the face recognition mode designed
in a landscape mode. As noticed, there is a camera
frame in the middle of interface for the purpose of face
recognition. Any face detected will then converted
into grayscale image and displayed at the top right
corner. So basically, the video frame provides the
program with lots of static image. This feature
provides more odds to be recognized by the program.
Once the student has been recognized by the program,
a text will be prompted to notify the user their
attendance has been marked. Student also has the
authority to access face image gallery. However, they
cannot add any new image.
J. Attendance log
Fig.10. Attendance log
Students can navigate to this windows activity by click
the “check attendance” button in the main menu.
Later, they will need to click again the “check
attendance” button in the current interface to see the
student attendance record appear in the location. This
button will also trigger the connection between
application and database. Thus, allowing attendance
record to be display. Notice that the interface is
divided into 2 sections and both section are scrollable
to fit all the information retrieved. The upper part is to
list down the entire student along with their details.
Meanwhile at the bottom, user can click on the student
name to display student’s location in the map and also
can print reports of attendance as fig. 11 shows an
example.
Fig.11. Database attendance log report
CONCLUSION
This paper proposed a smart, location-based time and
attendance tracking system that runs as a mobile
application on a smartphoneand uses location and face
detection as its core components.The classroom area is
set for tracking using GPS, and student coordinates
inside the area show thatthe student is present in the
class. The attendance system has been designed to
improve the efficiency of the student
attendance-taking process and to reduce the rate of
errors in managing students’ attendance records.
REFERENCES
[1] U. Jain, M. Shirodkar, V. Sinha and B. Nemade., “
Automated Attendance Management System using Face
Recognition,” IJCA Proceedings on International
Conference and Workshop on Emerging Trends in
Technology ICWET 2015(2):23-28, May 2015.
[2] A. Jha, “Class Room Attendance System Using Facial
Recognition System,” The International Journal of

International Journal of Advances in Electronics and Computer Science, ISSN: 2393-2835 Volume-3, Issue-8, Aug.-2016
Attendance System Using a Mobile Device: Face Recognition, GPS or Both?
32
Mathematics, Science, Technology and
Management,(ISSN : 2319-8125) Vol. 2 Issue 3.
[3] Lodha, R., Gupta, S., Jain, H. and Narula, H.,“Bluetooth
Smart Based Attendance Management System.”Procedia
Computer Science, 45, pp.524-527, 2015.
[4] 3M Electronic Safety Systems[Online]
Available on 4.5.16
athttp://www.3msys.com/products/biometric-attendance-sy
stems/
[5] Shoewu, O. O. M. Olaniyi, and Lawson, “Embedded
Computer-Based Lecture AttendanceManagement
System”, African Journal of Computing and ICT (Journal
of IEEE Nigeria ComputerSection), 4(3):27 – 36, 2011.
[6] Cheng, K., L. Xiang, T. Hirota, and K. Ushijimaa (2005),
“Effective Teaching for Large Classes withRental PCs by
Web System WTS”, Pro. Data Engineering Workshop
(DEWS2005), 1D – d3 (inJapanese).
[7] M. K. P. Basheer and C. V. Raghu, “Fingerprint
attendancesystem for classroom needs,” in Proc. India
Conference(INDICON), 2012 AnnualIEEE, pp. 433-438,
7-9 Dec. 2012.
[8] Shoewu, O. and O.A. Idowu (2012), “Development of
Attendance Management System usingBiometrics”, Pacific
Journal of Science and Technology, 13(1):300-307.
[9] Kadry, S., &Smaili, M. (2013). ”Wireless attendance
management system based on iris recognition”,Scientific
Research and essays, 5(12), 1428-1435.
[10] BISAM-BIS attendance Management System by
BISSoftware Development Services PVT Limited.
[Online].
Availableon 4.5.16 at
http://www.softwarehouse.co/school-attendancebrochure.
[11] RFIDSensNet Lab (2005), A white paper on
AutomaticAttendance System. Texas A & M University,
Texas, USA.
[12] Dubinsky, Y. and hazzan, O, “Improvement of Software
Quality: Introducing eXtreme Programming into a
Project-based Course. The proceeding of the 14th
International Conference of the Israel Society for Quality,
Jerusalem, Israel 2002.
