Ningthoujam Sunita Devi et al, / (IJCSIT) International Journal of Computer Science and Information Technologies, Vol.

5 (5) , 2014, 6491-6496

Face Recognition Using Principal Component

Analysis
Owiueyry6568391201-`=
Ningthoujam Sunita Devi1, K. Hemachandran2
1
Research Scholar, Department of Computer Science,
Assam University, Silchar, Assam,India,
2
Professors, Department of Computer Science,
Assam University, Silchar, Assam,India,

Abstract— Face recognition is one of the most relevant which was used for recognition of faces, to make it reliable.
applications of image analysis. It’s an efficient task (true For face identification the starting step involves extraction
challenge) to build an automated system with equal human of the relevant features from facial images. A big challenge
ability to face recognised. Face is a complex 3D visual model is how to quantize facial features so that a computer should
and developing a computational model for face recognition is
a difficult task. The paper presents a methodology for face
be able to identify a face. The study carried out by many
recognition based on information theory approach of coding researchers over the past several years indicates that certain
and decoding the face image. Proposed methodology is facial characteristics are used by human beings to identify
combination of two stages – Feature extraction using principle faces. Principle component analyses (PCA) is a classic tool
component analysis and recognition using the feed forward widely used in the appearance based approaches for
back propagation Neural Network. The proposed method has dimensionality reduction and feature extraction in most of
been tested on Oracle Research Laboratory (ORL) face the pattern recognition application. Hau T. Ngo et al.
database containing 400 images (40 classes). A recognition described a flexible and efficient architecture for real-time
score for the test lot is calculated by considering almost all the face recognition system based on modular Principal
variants of feature extraction. The test results gave a
recognition rate of 99.50%.
component Analysis method in an environment of FPGA,
they showed that modular PCA improves the accuracy of
Keywords—Face recognition, Principal component analysis face Recognition when face images have varying
(PCA), Artificial Neural network (ANN), Eigenvector, and expression and illumination. The architecture was able to
Eigenfaces. perform face recognition in 11ms for a database with 1000
face images [1].Boualleg proposed a new hybrid method
1. INTRODUCTION for the face recognition by combining the neural networks
The face is the primary focus of attention and plays a major with the Principal Component Analysis [2]. Sajid I et al.
role in identification and establishing the uniqueness of a presented a High performance FPGA based Face
particular person from the rest of the human society. In recognition system, where they used fixed point technique
spite of so many faces in the human society, there is with software hardware co-design methodology which
remarkable ability of a human eye to recognized one face reduces cycle and provides the flexibility in
from another. A human can recognize thousands of faces face[3].Hossein Sahoolizadeh proposed a new face
learned throughout the lifetime and identify familiar faces recognition method based on PCA(principal Component
at a glance even after years apart. This ability of human eye Analysis) LDA(Linear Discriminant Analysis) and Neural
is quite effective, even though there are changes in the Network. The proposed method was tested on orl database
visual stimulus due to aging of a person, expression and for face. Experimental results on this database rabel the
change of looks due to glasses, beards or in hair style. effectiveness of the proposed method for face recognition
There are many approaches such as security purpose, credit with less misclassification in comparison with previous
card verification, criminal identification etc. where the methods [4]. Sathaporn Visakhasart presented new
identification of a face plays an important role. A slight multipipelined architecture for face recognition system on
recognition of a particular person will be much better, in FPGA. This architecture helps to reduce the recognition
these kind of fields, than not even recognizing at all. time through its pipeline process and also encourage the
Although it is clear that people are good at face reduction in hardware resources [5]. Karim has been
recognition, but it’s not obvious that a human brain can developed the image processing and recognized the faces
encoded or decoded for every face. Human face recognition using PCA based face recognition techniques. And also
has been studied for more than twenty years. Developing a implement based on MATLAB program to identify the face
suitable program which can be used digitally in recognizing using Indian database and face recognition data. Matching
a face is a quite challenging task, because human faces are unknown image with known image, different techniques
complex and are different from each other in every aspect. like sum of absolute difference (SAD), Sum of squared
So, developing such a kind of program is a difficult task in difference (SSD), normalized cross correlation (NCC) etc.
this digital world, which may involve earlier techniques, The performance of PCA-based face recognition system is

www.ijcsit.com 6491
 Ningthoujam Sunita Devi et al, / (IJCSIT) International Journal of Computer Science and Information Technologies, Vol. 5 (5) , 2014, 6491-6496

quite satisfactory [6]. Janarbek Matai et al. presented image.The Proposed method is connection of two stages-
FPGA-based Real-Time Face Recognition System in which feature extraction using Principal component analysis and
they design full face detection and recognition system on recognition using the feed forward back propagation neural
FPGA vertex 5. They used haar data for detection and network.
Eigen[7]. Mohod approach rate the face recognition In this paper an unsupervised pattern recognition scheme,
problem as an intrinsically two dimensional (2D) which is independent of excessive geometry and
recognition problem rather than requiring recovery of 3D computation is proposed for a face recognition system. The
geometry, proceeds advantage of the fact that faces are system is implemented based on Eigenfaces, PCA and
normally upright and thus may be described by a small set ANN. PCA for face recognition is based on the information
of 2D characteristic views [8]. Rala M. Ebied describe a theory approach in which the relevant information in a face
method of feature Extraction using PCA and Kernel-PCA image is extracted as efficiently as possible. Further ANN
for Face Recognition in which they investigates the was used for classification. Neural Network notion is used
nonlinear kernel function to improvement the principal because of its ability to learn from observed data.
component analysis (PCA) for feature extraction. The Advantage and disadvantage of PCA
experiments carried out to investigate the performance of The advantages of PCA are listed below:
Kernel-PCA by comparing it with the performance of the 1) Lack of redundancy of data given the orthogonal
PCA. Two kernel functions are used with the kernel-PCA, components.
polynomial and Gaussian functions, to check which one 2) Reduced the complexity in face images grouping with
achieved a better performance. The k-nearest neighbor the use of PCA
classifier with Euclidean distance is used to investigate the 3) Smaller database representation since only the trainee
performance of the Kernel-PCA and PCA for classification images are stored in the form of their projections on a
step [9]. Manal Abdullah study to optimize the time reduced basis.
complexity of PCA (Eigenface) that does not affects the 4) Noise reduction since the maximum variation basis is
recognition performance [10]. chosen and so the small variations in the back ground are
There are two basic methods for recognized the face. The ignored automatically.
first method is based on feature extraction vectors from the 5)2DPCA over 1DPCA is that the feature vector is now
basic parts of a face feature such as eyes, nose, mouth, and two-dimensional so the problem of dimensionality is
chin, with the assist of deformable templates and extensive greatly reduced.
mathematics. Then the key information from the basic parts
of face is gathered and converted into a feature vector. Table 1. The features of PCA are shown in the table below.
Yullie and Cohen [11] used deformable templates in
contour extraction of face images. Feature Principal component analysis
PCA manages the entire data for the principal
Another method is based on the information theory Discrimination
components analysis without taking into
concepts viz. principal component analysis (PCA) method. between classes
consideration the fundamental class structure.
In this method, information that best describes a face is PCA applications in the significant fields of
Applications
derived from the entire face image. Based on the Karhunen- criminal investigation are beneficial
Loeve developed in pattern recognition, Kirby and Sirovich Computation for
PCA does not require large computations
large datasets
[12,13] used principal component analysis to efficiently
The directions of the maximum discrimination are
represent the pictures of faces. Any face image could be Direction of not the same as the directions of maximum
around reconstructed by a small collection of weights for maximum variance as it is not required to utilize the class
each face and a standard face picture, that is, eigen picture. discrimination information such as the within class scatter and
The weights are obtained by projecting the face image onto between class scatter
PCA examines the directions that have widest
the eigen picture. In mathematics, Eigenfaces are the set of Focus
variations
eigenvectors which are the set of feature vector or Supervised learning
PCA is an unsupervised technique.
characteristic used in the computer vision problem of technique
human face recognition. The principal component of the Well distributed
distribution of faces or the eigenvectors of the covariance classes in small PCA is not as powerful as other methods.
datasets
matrix of the set of face image is the Eigenfaces. Each face
can be represented exactly by a linear combination of the
Eigenfaces [14]. The best M eigenfaces construct an M Disadvantages of PCA are:
dimension (M-D) space that is called the “face space” 1) The covariance matrix is difficult to be evaluated in an
which is same as the image space. Turk and Pentland [15] accurate manner.
proposed a face recognition method based on the 2) The simplest invariance could not be captured even by
Eigenfaces approach. Gumus , Ergun [16] present an the PCA unless the training data explicitly provides this
evaluation of using various method for face recognition. information.
According to their experiment the classification accuracy 3) PCA is a less sensitive to different training data set.
increasing dimension of training data set, chosen feature 4) Computationally expensive and complex with the
extraction-classifier pairs. Agarwal,M., [17] present a increase in data size.
methodology for face recognition based on information 5) Time complexity is high.
theory approach of coding and decoding the face

www.ijcsit.com 6492
 Ningthoujam Sunita Devi et al, / (IJCSIT) International Journal of Computer Science and Information Technologies, Vol. 5 (5) , 2014, 6491-6496

2. PROPOSED TECHNIQUE image) and testing dataset (rest 40% images). The process
is described in Fig. 1.
The proposed technique is based on coding and decoding of
face images with emphasis on the significant of local and
global features of face. In this proposed method the
relevant information in a face image is feature extracted, Get the Face Descriptor Using Eigen Face
encoded and then compared with a face database of models The face library entries are normalized. Eigenfaces are
and then classified with ANN. The primary advantage of calculated from the training set and stored the data. An
the proposed method is independent of any judgment of number of face can be represented exactly in terms of a
features like open/closed eyes, different facial expressions linear combination of Eigenfaces. The face can also be
images, with and without Glasses. The model of the face approximated using only the best M eigen faces, that mean
recognition system shown in Fig 1: the largest eigen values. It interpretation for the most
variance within the set of face images. Best M Eigenfaces
span an M-dimensional subspace which is called the "face
space" of all possible images. For calculating the
Eigenfaces, the proposed PCA algorithm by Kirby and
Sirovich [18,17] was used. It includes the calculation of the
average face space and further the difference of average
with each face. The difference covariance matrix (C) is
used to compute for the dataset. The covariance matrix
between two sets of data reveals how much the sets
correlate. Based on the statistical technique called as PCA,
the number of eigenvector for covariance matrix can be
reduced from N (the number of pixels in image) to the
number of images in the training dataset.

FACE PARTICULAR PERSON 1-

n

Face feature Face feature Face feature Face feature

1 2 3 m

Fig. 1 – Face Library Formation and getting face

FACE DESCRIPTOR FOR TRAINING NEURAL NETWORK
Descriptor

3. PREPROCESSING AND FACE LIBRARY

FORMATION
Image size normalization, histogram equalization and
conversion into gray scale are used for preprocessing of the
face image. This process automatically reduce every face NEURAL NETWORK
image to X*Y pixels (based on user request) and distribute
the intensity of face images (histogram equalization) in
order to improve face recognition performance. In the
system Face images are stored in a face library. Every
exertion such as training set or Eigen face formation is OUTPUT
performed on this face library. The face library is further
divided into two sets – training dataset(60% of individual Fig. 2 – Training of Neural Network

www.ijcsit.com 6493
 Ningthoujam Sunita Devi et al, / (IJCSIT) International Journal of Computer Science and Information Technologies, Vol. 5 (5) , 2014, 6491-6496

Training of Neural Networks • Number of neurons in input layer : Number of

Only one ANN is used in which the no. of face descriptor eigenfaces to describe the faces
in the database are used an input to train the network. [19]. - Number of neurons in hidden
During training of the ANN’s, the faces signifier that Layer 10
belong to same person are used as positive examples for the - Number of neurons in output
person’s network (such that network gives 1 as output), and Layer 1.
negative examples for the different network. Fig. 2 shows • Transfer function of the ith layer: Tansig
schematic diagram for the networks training • Training Function: Trainlm
• Number of epochs used in training: 100
Simulation of ANN for Recognition • Performance function: mse
New test image is taken for recognition (from test dataset)
and set face images signifier is calculated from the
Eigenfaces (M) found before. These new signifier are given
as an input to one network; further these networks are
simulated. Contrast the simulated results and if the
maximum output exceeds the predefined approach level,
then it is assert that this new face belongs to the recognized
person with the maximum output (fig. 3) Fig.4: Network Architectur

NEW IMAGE PREPROCESSING Since the one network is equal to the number of people in
the database, therefore forty networks, one network was
created. Among the ten images, first 6 of them are used for
training the neural networks, then these networks are tested
and their properties are updated. The trained networks
NEW FACE PROJECTION ONTO EIGEN would be used later on for recognition purposes. For testing
DESCRIPTOR FACE the whole database, the faces used in training, testing and
recognition are changed and the recognition performance is
given for whole database. The complete face recognition
process is shown in Fig. 5.
NEURAL NETWORK DECISION CHOOSE
LARGEST VALUE

Fig. 3 – Testing of Neural Network

4. EXPERIMENT
The proposed method is tested on ORL face database.
Database has more than set of image of an individual’s face
with different conditions like expression, illumination etc.
There are ten different images of each of 40 perceptible
subjects. Each image has the size of 112 x 92 pixels with
256 grey levels. For some subjects, the images were taken
at assorted times, varying the lighting, facial expressions
(open / closed eyes, smiling / not smiling) and facial details
(glasses / no glasses). Each images were taken against a
dark homogeneous background with the subjects in an
upright, frontal position (with tolerance for some side
movement). A view image of the Face Database is
available (Fig. 5). The original pictures of 112 x 92 pixels
have been resized to 56 x 46 so that the input space has the
dimension of 2576.
Eigenfaces are calculated by using PCA algorithm and
experiment is performed by varying the number of
eigenfaces used in face space to calculate the face
descriptors of the images. The numbers of network used are
equal to number of subjects in the database.
The initial parameters of the Neural Network used in the
experiment are given below:
• Type: Feed forward back propagation network Fig. 5 – A complete process of PCA, Eigenface and ANN
• Number of layers: 3 (input, one hidden, based faced recognition system
output layer)

www.ijcsit.com 6494
 Ningthoujam Sunita Devi et al, / (IJCSIT) International Journal of Computer Science and Information Technologies, Vol. 5 (5) , 2014, 6491-6496

5. ANALYSIS TABLE 2: Recognition Score Of face recognition using

PCA and ANN (Testing)
The proposed technique is analyzed by varying the number No. of Network
MSE Percentage of classification
of eigenfaces used for feature extraction. The face Face Architecture
recognition performance is shown in Table I. The result 80 0.0344 89.3300
derived from proposed method is compared with the other 20 90 90 100 0.0289 93.120
techniques which are 1. 90 90 100 100 100 0.0201 94.135

K-means [20], 2. Fuzzy Ant with fuzzy C-means. [20] 80 0.1287 31.6533
Comparison of the result has been tabulated in Table 30 90 90 100 0.0214 92.6833
I.Table II: Recognition score of Face recognition using 90 90 100 100 100 0.0613 81.1167
PCA and ANN.
80 0.0742 71.1367

TABLE 1: Recognition Score Of face recognition using 40 90 90 100 0.0301 90.0833

PCA and ANN (Training) 90 90 100 100 100 0.0224 94.1067
80 0.1336 22.2867
No.Of Network Percentage of
MSE Iteration 50 90 90 100 0.0322 92.5033
Face Architecture classification
80 0.0573 87.5000 234 90 90 100 100 100 0.0248 94.2607
20 90 90 100 0.0160 99.500 208 80 0.1387 22.0667
90 90 100 100 100 0.0101 99.235 115 60 90 90 100 0.0460 90.6500
80 0.1287 34.5833 97 90 90 100 100 100 0.0325 91.0167
30 90 90 100 0.0214 99.5833 166 80 0.1306 31.4833
90 90 100 100 100 0.0613 87.9167 246 70 90 90 100 0.0470 88.0833
80 0.0742 79.1667 147 90 90 100 100 100 0.0352 92.6667
40 90 90 100 0.0301 97.0833 140 80 0.0612 86.6667

90 90 100 100 100 0.0224 99.1667 119 80 90 90 100 0.0675 81.333

80 0.1336 29.1667 74 90 90 100 100 100 0.0319 89.5667

50 90 90 100 0.0322 97.0833 196 80 0.0686 76.5867

90 90 100 100 100 0.0248 99.1667 148 90 90 90 100 0.0583 81.1833

80 0.1387 24.1667 85 90 90 100 100 100 0.0338 86.2500

60 90 90 100 0.0460 93.7500 95 80 0.1213 36.6667

90 90 100 100 100 0.0325 97.9167 150 100 90 90 100 0.0269 92.333

80 0.1306 34.5833 92 90 90 100 100 100 0.0435 85.7500

70 90 90 100 0.0470 92.0833 162

CONCLUSION
90 90 100 100 100 0.0352 96.6667 162 The paper presents a face recognition approach using PCA
80 0.0612 86.6667 117 and Neural Network techniques. The result is compared
80 90 90 100 0.0675 83.333 120 with several techniques and proposed technique gives a
better recognition rate then the other techniques. In the
90 90 100 100 100 0.0319 96.6667 155
Table.1 can see the recognition rate by varying the
80 0.0686 81.6667 117 eigenfaces and the maximum recognition rate obtained for
90 90 90 100 0.0583 87.0833 96 the whole dataset is 99.500. M Eigenfaces (‫ݑ‬௜ ) of highest
90 90 100 100 100 0.0338 96.2500 135
eigenvalue are actually needed to fabricate a complete basis
for the face space.
80 0.1213 41.6667 67 The Eigenfaces method is very sensitive to head
100 90 90 100 0.0269 98.333 122 orientations, and most of the distinct occur for the images
90 90 100 100 100 0.0435 93.7500 139 with large head orientations. By choosing PCA as a feature
selection technique (for the set of images from the ORL
Database of Faces), one can reduce the space dimension
from 2576 to 50 (equal to no. of selected eigenfaces of
highest eigenvalue).

www.ijcsit.com 6495
 Ningthoujam Sunita Devi et al, / (IJCSIT) International Journal of Computer Science and Information Technologies, Vol. 5 (5) , 2014, 6491-6496

REFERENCES faces." Pattern Analysis and Machine Intelligence, IEEE

1. Ngo. H. T, Gottumukkal. R , and Asari. V. K, “A flexible and Transactions on 12, no. 1 (1990): 103-108.
efficient hardware architecture for realtime face recognition based on 13. Sirovich, Lawrence, and Michael Kirby. "Low-dimensional
eigenface,” in IEEE Computer Society Annual Symposium on VLSI, procedure for the characterization of human faces." JOSA A 4, no. 3
2005. (1987): 519-524.
2. Boualleg. A. H., Bencheriet Ch, and Tebbikh. H, "Automatic Face 14. Moon, Hyeonjoon. "Biometrics person authentication using
recognition using neural network-PCA." In Information and projection-based face recognition system in verification scenario."
Communication Technologies, 2006. ICTTA'06. 2nd, vol. 1, pp. In Biometric Authentication, pp. 207-213. Springer Berlin
1920-1925. IEEE, 2006. Heidelberg, 2004.
3. Sajid. I., Ahmed M. M., Kresimir Delac, Mislav Grgic, Sonja 15. Turk, Matthew, and Alex Pentland. "Eigenfaces for
Grgic,” Independent Comparative Study of PCA, ICA, and LDAon recognition." Journal of cognitive neuroscience 3, no. 1 (1991): 71-
the I. Taj, M. Humayun, and F.Hameed,“Design of high performance 86.
fpga based face recognition system,” in PIERS Proceedings, 16. Gumus,Ergun,et.al., “ Evaluation of Face Recognition Techniques
Cambridge, USA, July 2-6, 2008, pp.504-510 Using PCA,Wavelet and SVM ”,Expert System with Application
4. Sahoolizadeh. A , Hossein.B , Zargham Heidari, and Hamid 37.9(2010) : 6404-6408.
Dehghani.C "A new face recognition method using PCA, LDA and 17. Agarwal, Mayank, Himanshu Agrawal, Nikunj Jain, and Manish
neural network." International Journal of Computer Science and Kumar. "Face recognition using principle component analysis,
Engineering 2, no. 4 (2008): 218-223. eigenface and neural network." In Signal Acquisition and
5. Sathaporn Visakhasart, Orachat Chitsobhuk, “Multi- Pipeline Processing, 2010. ICSAP'10. International Conference on, pp. 310-
architecture for face recognition on FPGA,”International Conference 314. IEEE, 2010.
on digital Image Processing, 2009, pp.152-156. 18. Kirby, Michael, and Lawrence Sirovich. "Application of the
6. Karim,T.F,Lipu,M.S.H, Rahman, M.L., & Sultana ,F.Face Karhunen-Loeve procedure for the characterization of human
Recognition Using PCA-based method. In advanced Management faces." Pattern Analysis and Machine Intelligence, IEEE
Science (ICAMS) 2010.IEEE International Conference on Transactions on 12, no. 1 (1990): 103-108.
(Vol.3,pp.158-162). 19. Victor-Emil and Luliana-Florentina, “Face Recognition using a fuzzy
7. Janarbek Matai, Ali Irturk and Ryan Kastner,” Design and – Gaussian ANN”, IEEE 2002. Proceedings, Aug. 2002 Page(s):361
Implementation of an FPGA-based Real-Time Face Recognition – 368.
System”, in IEEE International Symposium on Field Programmable 20. Wijaya.S.L, M. Savvides, and B. V. K. V. Kumar, "Illumination-
Custom Computing Machines, 2011, pp 97-100. tolerant face verification of low-bitrate JPEG2000 wavelet images
8. Mohod, Prakash S., and Kalpna C. Jondhale. "Face Recognition with advanced correlation filters for handheld devices," Applied
Using PCA. "International Journal of Artificial Intelligence and Optics, Vol.44, pp.655-665, 2005.
Knowledge Discovery 1, no. 1 (2011): 25-28. 21. Gong, Shaogang, Stephen J. McKenna, and Alexandra
9. Rala M. Ebied,” Feature Extraction using PCA and Kernel-PCA for Psarrou. Dynamic vision: from images to face recognition. Imperial
Face Recognition”, in The 8th International Conference on College Press, 2000.
INFOrmatics and Systems Computational Intelligence and 22. Acosta, Emiliano, Luis Torres, Alberto Albiol, and Edward Delp.
Multimedia Computing Track , 2012 ,pp mm72-mm77 "An automatic face detection and recognition system for video
10. Abdullah, Manal, Majda Wazzan, and Sahar Bo-saeed. "Optimizing indexing applications." In Acoustics, Speech, and Signal Processing
Face Recognition Using PCA." arXiv preprint (ICASSP), 2002 IEEE International Conference on, vol. 4, pp. IV-
arXiv:1206.1515 (2012). 3644. IEEE, 2002.
11. Yuille, Alan L., Peter W. Hallinan, and David S. Cohen. "Feature 23. Manjunath, B. S., Rama Chellappa, and Christoph von der Malsburg.
extraction from faces using deformable templates." International "A feature based approach to face recognition." In Computer Vision
journal of computer vision 8, no. 2 (1992): 99-111. and Pattern Recognition, 1992. Proceedings CVPR'92., 1992 IEEE
12. Kirby, Michael, and Lawrence Sirovich. "Application of the Computer Society Conference on, pp. 373-378. IEEE, 1992.
Karhunen-Loeve procedure for the characterization of human 24. Demuth, Howard, and Mark Beale. "Neural network toolbox for use
with MATLAB." (1993).

www.ijcsit.com 6496