Content uploaded by Etinosa Noma-Osaghae
Author content
All content in this area was uploaded by Etinosa Noma-Osaghae on Jul 27, 2018
Content may be subject to copyright.
A Face Recognition Attendance System with GSM
Notification
Kennedy Okokpujie1, Etinosa Noma-Osaghae1, Samuel John1, Kalu-Anyah Grace1, Imhade Okokpujie2
1Department of Electrical and Information Engineering
2Department of Mechanical Engineering
Covenant University, Ota
Ogun State, Nigeria
kennedy.okokpujie, etinosa.noma-osaghae, samuel.john, Imhade.okokpujie (@covenantuniversity.edu.ng)
Abstract — Current biometric methods for attendance are too
intrusive. This paper presents a stress-free non-intrusive way of
taking class attendance using face as the biometric. It also has the
added novelty of relaying vital information about class attendance
to handheld devices via any available cellular network. During
enrollment, a camera was used to acquire facial images that were
made into templates using Fisherfaces algorithm. These templates
were stored in a database. During verification or attendance taking,
facial features extracted from acquired face images and stored
picture templates were compared using Fisher Linear
Discrimination algorithm for any match within the pre-set
threshold. Vital information about collated attendance reports were
sent via a cellular network to designated handheld devices. The
designed and implemented system had 54.17% accuracy during
verification when lighting was varied without any variation in
facial expression during enrollment. The system had 70.83%
accuracy during verification when facial expressions were varied
along with variations in lighting conditions during enrollment.
Keywords — Fisherfaces; feature extraction; enrollment;
verification; discrimination; facial recognition; cellular network
I. INTRODUCTION
Biometrics has an epic range of applications and more
innovative ways of using it keeps emerging. A pivotal feature
of biometrics is its relative permanence. All biometrics are
distinctively unique in features and in individuals [16].
Biometrics has been around for more than a century, with
the earliest systematic capture of a hand image for
identification purposes recorded in 1858 [1]. In the following
years, other methods began to emerge. Including iris pattern
recognition, face recognition, speech recognition, signature
analysis, gait pattern, palm prints, among others. Biometrics
has come a long way since the late nineteenth century and is
certainly a constantly developing area of technology [2]. The
applications of biometrics are vast and new applications are
emerging.
Typically, a good biometric system has two phases; the
enrollment phase and the recognition phase. Enrollment
involves obtaining the biometric trait of an individual, storing
the features in a database as well as an identifier to enable the
trait to be associated with the individual. The recognition phase
involves acquiring the biometric trait, extracting the identifier
and checking the database to see if there is any match [8].
Face recognition has numerous merits above other
biometric methods. Most of the other biometric forms need
some form of action by the user. However, face recognition can
be done without the involvement of the user due to the fact that
face images can be acquired from a distance by a camera. In
the end, face recognition is totally non-intrusive and so does
not expose the user to germs that may be prevalent in a system
that has multiple users. Face recognition is non-intrusive [7].
The major aim of this paper is to implement a face
recognition attendance system that has the added novelty of
relaying outcomes of an attendance taken via a cellular
network to designated mobile devices. It also gives a peak into
how variations in lighting, facial expression and angles affects
the accuracy of the designed and implemented face recognition
attendance system.
The objectives of this paper are to design a face recognition
attendance system backed by a database, implement the design,
add the novelty of being able to send results of attendance
taken via a cellular network to designated handheld devices
and test to see how variations in face angle, facial expression
and lighting affects the accuracy of the designed and
implemented face recognition attendance system.
A light-variation, facial expression resistant face
recognition attendance system with Cellular network
information relay emerged. This could potentially serve as part
of a home automation system that notifies home owners of the
identity of persons present in their homes regardless of where
they are. It can also serve as a deterrent to intentional class
absentees whose parents would be notified by the system.
Limitations to this study include partial covering of the face
because of hairstyle and glasses, too low or too high image
quality that could make analysis more complex or slow down
the computer processor’s speed respectively. A balance was
maintained that did not distort expected results.
I. REVIEW OF SIMILAR WORKS
Akshay N. Patil et al [11] presented the design of a system
aimed at providing essential security to homes, banks and for
similarly-related control operations and to send security alerts
utilising the GSM module. A Raspberry Pi module was used to
operate and control the video camera (by capturing images)
and turn on a relay that unlocks the door. Six images were
captured and used to create a database. The Raspberry Pi
module coordinated the comparison between stored images in
the database and captured live images. The result of the
comparison was sent to the GSM module which in turn relayed
the result via a cellular network to the handheld devices of
designated persons. The result of the comparison was either
positive or negative. If the result was positive, the message sent
was “Person Identified!! DOOR OPENS!!" otherwise the
message sent was “unknown person is trying to unlock the
door”.
Ashish Choudhary et al [12] proposed a system that solves
problems and limitations associated with conventional methods
of attendance capture by automating the process of attendance
management during exams and lectures thus saving effort and
time. The system made use of a camera that processes videos in
real time. The faces of students attending a class were captured
and stored in a database. The information stored in the database
was then used to take attendance. To make the system fool
proof, attendance was taken twice, once at the beginning of the
class and at the end of the class. Data about face images taken
at the beginning and end of the class were used to decide
whether a student was present or absent for the day. A student
is marked present only if he is recognised in both the images.
Bhange Sourabh Sanjay et al [13] proposed a vehicle
security system. As soon as a vehicle is started the identity of
the driver is ascertained. This system makes sure only
accredited users are allowed to drive the car. If an unauthorised
person is identified or an unknown person, the system would
prevent that person from driving. The location of the car and an
image of the unauthorised person are sent to the system
controller which has the ability to prevent the vehicle from
moving further.
Sarvesh V. A. et al [14] proposed a system to offer
advanced security program in cars, the system was to be
enabled with a custom password and a face recognition
program, a GSM module and a control platform. The program
is mainly used to notice the thief who is trying to steal the car.
The FRS (Face Recognition System) is used to detect the face
of the driver and recognise it. The GSM plays a very important
part in this system. The prototype used a Microcontroller and
GSM service [15].
II. RELATED STUDIES
A. Face Detection Technology
The computer application in conjunction with the required
hardware must be able detect the presence or absence of a face
in any digital image. The resource of any worthy face
recognition system must be directed at the face and be able to
tell if a face is absent in a digital image. The main means of
achieving this is through the process of feature extraction.
There are several means of extracting features from a face and
some of them are:
Feature-based method: the focus here is on texture and
skin colour [1]. It pays no real consideration position,
viewpoints and lighting conditions.
Knowledge based approach: the face structure is made
into a set of governing rules that is followed by the
algorithm performing feature extraction on the face. It
was made popular by researchers and their deep
knowledge about the structure of the human face. [2].
Appearance based approach: template models are used
to extract the features of a face in this approach [2].
Template matching: this method uses patterns that have
been designed to describe the face. The input image is
compared to the designed pattern and used create the
needed face template. This approach is highly affected
by variations in scale and pose.
B. Viola Jones Algorithm
The face detection algorithm utilised was the Viola-Jones
algorithm, an appearance based method for detecting faces. It
detects faces speedily and quite reliably [17]. The processes
used by the algorithm to detect faces are detailed below:
Haar Feature Selection: similar features shared by all
human faces are used to make the rectangular training
pattern used to get a match when trying to detect a face
[3].
Integral image: the integral of images captured by the
camera are computed to enable fast detection of
whether a face is present or absent in an image [1].
AdaBoost Algorithm: Adaptive boosting is the process
of taking a few feature from a large pool of potential
features in an image that may contain a face [1][4][5].
Attentional Cascade: here attention is directed towards
promising parts of an image that may contain a face by
using more complex classifiers in a cascade structure
[4].
C. Face Recognition Technology
There are several methods used by existing technologies to
recognize faces and some of them are:
Holistic Approach: here, all regions of the face are
considered when trying to recognise a face. It is a
comprehensive approach to face recognition [6].
Feature-Based Approach: here, only a few unique
features are used to recognize a face [7][18][19].
Appearance based approach: template models are used
to extract the features of a face in this approach [2].
Hybrid Approach: here, combinations of distinct and
comprehensive approach are used to recognized faces
[8].
D. Fisherfaces
To obtain the template of faces captured, the Fisherface
primary goal is to apply the Fisher Linear Discriminant (FLD)
[9].
E. GSM
GSM is an acronym for Global System for Mobile
Communications. It was introduced in the 1990s as a digital
technology based on Circuit Switching and Cellular concept. It
is a form of mobile communications [10].
III. METHODOLOGY
A camera served as the input device. During enrollment,
Viola-Jones algorithm was used to detect faces and the camera
acquired the detected faces. Fisherfaces algorithm was used to
create templates of the faces that were captured. A database
stored the created templates along with other particulars unique
to the users enrolled. During verification, the camera acquired
images of faces detected. The acquired images were compared
with face templates stored in the database for any match. The
number of verified face images formed the basis for attendance
taking by the attendance algorithm. Information about
attendance taken was relayed via a cellular network to
authorized handheld devices. The algorithms used were
implemented using C++. The created database had a capacity
of twenty (20) persons. Each person had as least Twenty (20)
and at most Fifty (50) images of their face stored in the
database.
A. Componets
The main parts of the designed system are:
1) Hardware:
a) A Webcam.
b) A Modem.
c) A Personal Computer.
2) Software:
a) PostgreSQL was used to create the database.
b) Qt Software Development Kit (SDK) was used to
create the graphical user interface application that was used to
interact with the database, camera and cellular phone. The
graphical user interface created allowed an administrator to
access the system with a username and password, enroll users
and update user bio-data. For user authentication, the system
was accessed using the same username and password.
c) OpenCV library provided the C++ Fisherfaces
algoritm.
Fig.1. Administrator’s Login Window.
Fig. 2. Sample Face Image
B. Use Case Diagram
The interaction between the user and the software
application is shown diagrammatically in Figure 3. The case
diagram shows the set of actions that were performed by the
system and how it interacts with all users of the system. It
displays the different purposes of the system and also relates
what the system is capable of and lays the rules of interaction
for the required service. It can be referred to as the blueprint of
the system.
C. Database
The database held the information of twenty individuals,
stored under columns indicating title, surname, middle name,
gender, date of birth, nationality, state of origin and local
government area. Each entry had provision for the storage of
images, with a minimum of twenty and maximum of fifty
images per person. The type of database used to implement this
system was relational. The interface for enrolling students’ bio
data is shown in Figure 4 and Figure 5.
Fig. 3. Use Case Diagram
D. User Identification Portal
After enrollment, verification was carried out through the
user identification portal and when a match is found the user is
marked present for the day.
E. Information Relay System
After attendance was taken, a set of codes enabled the
automatic extraction of vital information about attendance
from the database. The application portal relayed the
information got to the modem. The modem then transfers the
information through a cellular network to the designated
handheld device.
Fig. 4. Bio Data Table
Fig. 5. Application Database for Enrolled Students
IV. RESULT AND DISCUSSION
A. Testing
To determine the accuracy of the recognition system, tests
were carried out:
1) Under various lighting conditions and varying facial
expressions.
2) By varying facial expressions and angles along with
lighting conditions.
3) The tests were carried out on the particulars of twelve
individuals stored in the database.
4) The identification process was carried out ten (10)times
to obtain the probability of an accurate match.
The table below indicates the number of times an
individual was identified correctly [Number of Positives
(NOP)], and the number of times there was a false negative
match [Number of False Negative Match (NFN)].
Table I. Results from Tests carried out by Varying Facial
Expressions and Angles along with the Lighting Conditions.
S/
N
Name
NOP
NFN
Accuracy
1
Jane
7
3
70
2
Grace
6
4
60
3
Nkechi
7
3
70
4
Damilola
6
4
60
5
Gift
9
1
90
6
Uche
8
2
80
7
Chiamaka
8
2
80
8
Temilola
7
3
70
9
Jonathan
6
4
60
10
Chisom
6
4
60
11
Ajulibe
7
3
70
12
Kusimo
8
2
80
The average accuracy (by varying facial expressions
and angles along with the lighting conditions):
Table II. Results from tests carried out under various lighting
conditions but unvarying facial expressions.
S/N
Name
NOP
NFN
Accuracy
1
Jane
5
5
50
2
Grace
4
6
40
3
Nkechi
5
5
50
4
Damilola
5
5
50
5
Gift
7
3
70
6
Uche
6
4
60
7
Chiamaka
8
2
80
8
Temilola
4
6
40
9
Jonathan
3
7
30
10
Chisom
6
4
60
11
Ajulibe
7
3
70
12
Kusimo
5
5
50
The average accuracy of the system (under various
lighting conditions but unvarying facial expressions):
The results showed that the system responds better to
face expression variation than to lighting variation.
CONCLUSION
The designed and implemented face recognition
system worked with varying levels of accuracy. A
combination of lighting, facial and angular factors were
responsible for the variations in accuracies got from the tests
carried out on the implemented design.
Results obtained showed clearly that the face
recognition attendance system performs better in terms of
accuracy when facial expressions and angles are varied along
with lighting conditions during enrollment (at least twenty
(20) face images for each enrollee).
As a requirement, the designed face recognition
attendance system is used only under good lighting conditions.
FUTURE WORK
1) Embedded systems would be incorporated into the
attendance system to make it fully autonomous.
2) The system would be adapted for transmission of
attendance information via local area networks and WIFI to
designated nodes on a network.
ACKNOWLEDGMENT
This paper was sponsored by Covenant University,
Ota, Ogun State, Nigeria.
REFERENCES
[1] G. Eason, B. Noble, and I.N. Sneddon, “On certain integrals of
Lipschitz-Hankel type involving products of Bessel functions,” Phil.
Trans. Roy. Soc. London, vol. A247, pp. 529-551, April 1955.
[2] J. Clerk Maxwell, A Treatise on Electricity and Magnetism, 3rd ed., vol.
2. Oxford: Clarendon, 1892, pp.68-73.
[3] I.S. Jacobs and C.P. Bean, “Fine particles, thin films and exchange
anisotropy,” in Magnetism, vol. III, G.T. Rado and H. Suhl, Eds. New
York: Academic, 1963, pp. 271-350.
[4] K. Elissa, “Title of paper if known,” unpublished.
[5] R. Nicole, “Title of paper with only first word capitalized,” J. Name
Stand. Abbrev., in press.
[6] Y. Yorozu, M. Hirano, K. Oka, and Y. Tagawa, “Electron spectroscopy
studies on magneto-optical media and plastic substrate interface,” IEEE
Transl. J. Magn. Japan, vol. 2, pp. 740-741, August 1987 [Digests 9th
Annual Conf. Magnetics Japan, p. 301, 1982].
[7] M. Young, “The Technical Writer’s Handbook”. Mill Valley, CA:
University Science, 1989.
[8] W. Zhao, R. Chellappa, P.J. Phillips, and , A. Rosenfeld. “Face
recognition: A literature survey”. ACM computing surveys (CSUR)
2003, 35(4), pp.399-458.
[9] P. N. Belhumeur, J.P. Hespanha, and D.J. Kriegman. “Eigenfaces vs.
fisherfaces: Recognition using class specific linear projection”. IEEE
Transactions on pattern analysis and machine intelligence, 1997 19(7),
pp.711-720.
[10] K.O. Okokpujie, O. O. Uduehi, F. O. Edeko. “An Innovative Technique
in ATM Security: An Enhanced Biometric ATM with GSM Feedback
Mechanism” Journal of Electrical and Electronics Engineering (JEEE)
2015, 12 (2), Pages 68-81
[11] A. N. Patil, R. B. Ranavare, D.V. Ballal, and A.P.P.Kotekar. Raspberry
“Pi Based Face Recognition System For Door Unlocking”. unpublished.
[12] A. Choudhary, A. Tripathi, A.Bajaj, M. Rathi and B. M. Nandini.
“Attendance Generation Using Facial Recognition”. unpublished.
[13] S. S Ghodke, H. Kolhe, S. Chaudhari, K. Deshpande, and S. Athavle, S.
“ATM transaction security system using biometric palm print
recognition and transaction confirmation system”. International Journal
Of Engineering And Computer Science,2014, 3(4), pp.5332-5335
[14] V. A. Sarvesh, R. Chetana. “Face Recognition System for Unlocking
Automobiles Using GSM and Embedded Technology”. International
Journal of Advanced Research in Electrical, Electronics and
Instrumentation Engineering Vol. 5, Issue 7, July2016.pp 6090-6097
[15] K. Okokpujie, F. Olajide, S. John and C. G. Kennedy. “Implementation
of the Enhanced Fingerprint Authentication in the ATM System Using
ATmega128”. In Proceedings of the International Conference on
Security and Management (SAM 2016) (p. 258). The Steering
Committee of The World Congress in Computer Science, Computer
Engineering and Applied Computing (WorldComp).
[16] T. O. Majekodunmi, and F.E. Idachaba. “A review of the fingerprint,
speaker recognition, face recognition and iris recognition based
biometric identification technologies”.
[17] R. Jafri, and H. R. Arabnia. “A survey of face recognition techniques”.
Jips,2009 5(2), pp.41-68.
[18] Okokpujie, K., Etinosa, N.O., John, S. and Joy, E., 2017, November.
Comparative Analysis of Fingerprint Preprocessing Algorithms for
Electronic Voting Processes. In International Conference on Information
Theoretic Security (pp. 212-219). Springer, Singapore.
[19] Okokpujie, K., Etinosa, N.O., John, S. and Joy, E., 2017, November.
Comparative Analysis of Fingerprint Preprocessing Algorithms for
Electronic Voting Processes. In International Conference on Information
Theoretic Security (pp. 212-219). Springer, Singapore.