You are on page 1of 7

International Journal of Innovations & Advancement in Computer Science

IJIACS
ISSN 2347 – 8616
Volume 6, Issue 11
November 2017

Face Recognition Based On Eigen Vectors

B.Grace Jemimah1, K.Rajendra Babu2, K.Kamar Taz3, C.Nagaraju4


academic cunsultent1, scholar2, scholar3,Assoc.proffessor4
Department of Computer Science and Engineering
YSR Engineering College of Yogi Vemana University, Proddatur

Abstract: Face recognition plays an important role in human and machine interaction. To increase the recognition, rate
many methods have been implemented. Among those Principle Component Analysis is one of the good methods.
Obtaining dissimilarity of face feature is an important feature in this method, however it is sensitive to scaling factor. To
overcome this the linear discriminant Analysis is suitable for invariant scaling factor, however it is sensitive to high
noise. To overcome this, we proposed Eigen vectors to invariant scaling and less noise. In this method after calculating
the mean we take the values of Eigen Vectors and biggest Eigen value is considered. Then features of the face will be
extracted clearly than the earlier methods.
Keywords: PCA, LDA, EIGENVECTORS, EIGEN VALUES

1. Introduction:
In face recognition, the basic step is the feature extraction followed by a classifier. Facial feature extraction
can be performed in two ways that is either by geometric feature based method or by appearance based
method. The former methods need accurate detection and tracking of facial feature points [1,2 In [3] proposed
how to face recognition based on PCA, multi class classification is used to get highest value of log. In [4] is
the processes that involves the study of facial features, recognizing and comparing with the images in the
database. In [5] explained about how the light exposure impact will affect the accuracy of the face. In [6] the
expressions of the face are recognized by using PCA which is further eigenvectors are applied to it, so that we
can easily know what the expression is presented in the face. In [7] the process orientation and the lightening
conditions in the image will be recognized, so that the orientation will be different. In [8] PCA is used to
recognize the face and later shortest distance between the faces will calculated by using the Linear
Discriminant Analysis(LDA).as to increase the face recognition rate. In [9] the PCA is implemented based on
the mean and finding the normalization by subtracting the mean from the data set.so that common features of
the data will be removed. In this method we use the median instead of the mean to calculate the central
tendency of the normalization. In [10] PCA, Eigen face and vector machines are used to recognize the face.
By using N-fold method the face will be divided into two sections and from the vector values the features of
the faces will be extracted. In [11] the PCA is represented as weighted Eigen vectors. The recognition is done
by allowing the image on the subspace covered by the Eigen faces and Euclidean distance is calculated. In
[12] face recognition is explained by showing that PCA and LDA as the parameters and shown that LDA is
better than the PCA. In [13] histogram equalization is used to detect the color images in the different
lightening conditions. In [14] dimensionality of the image will be reduced and intra-class discriminating
characteristics will be calculated. In [15] explained how to recognize the color image and convert into the gray
scale image and extract the features of the face. In this paper, we propose a complete face recognition system
by combining the face tracking techniques. The system automatically detects and extracts the human face
from the background is based on a combination of Eigen values. The faces are analyzed in the spatial and the
frequency domains. These faces consist of the patterns which are easily recognized at higher levels

2.Principal Component Analysis (PCA)


The Principle Component Analysis (PCA) is one of the techniques that has been used in compression and
image recognition. PCA can do data compression, prediction, feature extraction redundancy removal etc. For
maximum variance, data set of observation is { } where n=N, is Euclidean value with dimensionality D.

674 B.Grace Jemimah, K.Rajendra Babu, K.Kamar Taz, C.Nagaraju


International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 6, Issue 11
November 2017

Dimensionality M<D is considered, the variance of considered data is maximized and the value of M is
assumed.
The mean of the projected data is given by:
Y= ∑
The variance of projected data is:
1
{ − } =

The data covariance matrix S is represented as:


S= ∑ ( − )( − )

Fig1: Face recognition using PCA


Disadvantages
 The covariance matrix is difficult to solve accurately and hence is time consuming [16].
 The size and location of each face image must remain similar [17].
 The method is very sensitive to scale.
3.Linear Discriminant analysis (LDA)
Every face feature consists of unique discrimination power under person’s gender, race and the age. The LDA
is used to find the linear transformations which extracts the features with scatter matrix analysis. The ratio of
the number of the test sample to the total number of test sample is defined as follows:
:
=
:
Now we have to subtract this maximum value from all the other samples in the matrix:
1= −
The covariance matrix will be calculated so that we can get the values
( )
=

Discriminant function is as follows:


1
=µ − µ µ + log ( )
2

Fig 2: face recognition using LDA

675 B.Grace Jemimah, K.Rajendra Babu, K.Kamar Taz, C.Nagaraju


International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 6, Issue 11
November 2017

4.EIGEN VECTORS
This method is the simplest and best approach for face recognition. This approach transforms image into a
small set of essential characteristics, Eigen faces which are most important for the initial set of images i.e.
trained image [18]. The first step that we want to do is calculate the sum and the global mean of the elements
in the matrix.
= 1/ (∑ )
Subtract the mean from each element in the matrix
Ф= −
Calculate the covariance matrix
T=CCB
Where T = {Ф1, Ф2,……. ФM}.
The Eigen vectors and Eigen value of C such that
T =
Where = Eigen vector
=Eigen value
Subtract Eigen vector from Eigen values and assign to
= −
Get the transpose of
=( )
Finally, multiply with
= ×

Fig 3: Face recognition using Eigen vectors

5. Experimental Results:
The performance of the Eigen Vectors technique was done experimentally on different databases and
compared with PCA and LDA. Seven parameters like mean, standard-deviation, logarithmic-values, hist
values, jaccord formula, PSNR values and chi-square values are evaluated on ten different databases with each
seven face images. This technique gives better outcomes on ORL and YALEB, CK, FERET, LFW, CAS-
FEAL, FIE databases and the database created by me for my DST funding projects. Minimum value in mean
and chi-square graphs and tables specifies the recognition of query image and remaining graphs and tables
show the maximum value specifying recognition of query image.

676 B.Grace Jemimah, K.Rajendra Babu, K.Kamar Taz, C.Nagaraju


International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 6, Issue 11
November 2017

Database Images

PrincipalComponentAnalysis

Linear Discriminant Analysis

EigenFaces

677 B.Grace Jemimah, K.Rajendra Babu, K.Kamar Taz, C.Nagaraju


International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 6, Issue 11
November 2017

Table1:hist Table2:log Table: chi-square Table: mean

Table5: STDV Table6: JACCORD Table7: PSNR

Graphs:
HIST LOG

pca eig lda pca eig lda

150 40
100
20
50
0 0
1 3 5 7 9 11 13 1 3 5 7 9 11 13

HIST LOG
CHI MEAN

pca eig lda mean pca mean eig


100 mean lda
20
50
0 0
1 3 5 7 9 11 13 1 3 5 7 9 11 13

CHI MEAN

678 B.Grace Jemimah, K.Rajendra Babu, K.Kamar Taz, C.Nagaraju


International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 6, Issue 11
November 2017

STDV JACCORD
jaccord pca jaccord eig
stdv pca stdv eig jaccord lda
stdv lda 200
500

0
0
1 3 5 7 9 11 13
1 3 5 7 9 11 13
STDV JACCORD

PSNR
PSNR pca PSNR eig
PSNR lda
20

0
1 3 5 7 9 11 13

PSNR

6.Conclusions:
In this paper, Eigen Vectors technique is proposed to detect and recognize the face. Mean and biggest Eigen
values is used to get the features of the face in different environmental conditions. This method is compared
with Principle Component Analysis and Linear Discriminant Analysis. The Eigen Vectors is better to detect
and recognize the face in constrained and unconstrained conditions than existing techniques. However, this
technique fails to detect the small features like moles and any other scars on the face which doesn't playmuch
role in the face recognition.

7. References:
[1] L. Zhang and D. Tjondronegoro, "Facial Expression Recognition Using Facial Movement Features,"ITAC, vol.2(4),
pp.219–229, 2011.
[2] A.R. Rivera, J.R. Castillo and O. Chae, "Local Directional Number Pattern for Face Analysis: Face and Expression
Recognition",ITIP, vol.22(5), pp.1740–1752, 2013.
[3] YaminiPiparsaniyan, Vijay K. Sharma, K. K. Mahapatra, "RobustFacial Expression Recognition using Gabor Feature
and Bayesian Discriminating Classifier"
[4]Gurpreet Kaur, Sukhvir Kaur, Amit Walia, "Face Recognition using PCA,Deep Face Method" IJCSMC, Vol. 5, Issue.
5, May 2016, pg.359 – 366
[5] Suad Haji, AsafVarol, RTFRS,ISDFS,25-27 April 2016,LittleRock,AR
[6] Rehmat Khan, Rohit Raja, "Neural Network associated with recognition of Facial Expressions of Basic
Emotions"IJCTT volume3Issue4- 2012
[7] Vinay Rishiwa, Ashutosh Gupta, "Improved PCA Algorithm for Face Recognition" WAP, Vol (2), Issue (1), January
2012. 55-59
[8] Sukhvinder Singh, Meenakshi Sharma and Dr. N Suresh Rao, "Accurate Face Recognition Using PCA and LDA",
ICETCIP, Bangkok Dec., 2011
[9] G. Shree Devi, M. MunirAhamed Rabbani, A. Jaya "Face recognition using principal component analysis with
median for normalization on a heterogeneous data set".
[10] ParvanehShayghanGharamaleki and HadiSeyedarabi "Face recognition using Eigen faces, PCA and support vector
machines", EJAESR, 2015, 4 (3):24-30
[11] Liton Chandra Paul, Abdulla Al Sumam, "Face Recognition Using Principal Component Analysis
Method",IJARCET Volume 1, Issue 9, November 2012

679 B.Grace Jemimah, K.Rajendra Babu, K.Kamar Taz, C.Nagaraju


International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 6, Issue 11
November 2017

[12] S. K. Hese, M. R. Banwaskar,"Performance Evaluation of PCA and LDA for Face Recognition",IJARECE, Vol 2,
Issue 2, February 2013
[13] Parteek Kumar,Praveen Sehgal "Face detection using principal component analysis", IJCET,Volume 7, Issue 3,
May-June 2016, pp. 174–178, Article ID: IJCET_07_03_016
[14] S. Thakur, J. K. Sing, D. K. Basu, M. Nasipuri,M.Kundu,"Face Recognition using Principal Component Analysis
and RBF Neural Networks" S. THAKUR et al.
[15] C.Nagaraju, B.Srinu, E. Srinivasa Rao,” An efficient Facial Features extraction Technique for Face Recognition
system Using Local Binary Patterns”, (IJITEE) ISSN: 2278-3075, Volume-2, Issue-6, May 2013.
[16] P. J. Phillips, P. J. Flynn, T. Scruggs, K. W. Bowyer, J. Chang, K. Hoffman, J. Marques, J. Min and W. Worek,
"Overview of the Face Recognition Grand Challenge, in Computer vision and pattern recognition", 2005. CVPR
2005. IEEE Computer Society Conference on, 2005, pp. 947–954.
[17] S.K. Singh, D. S. Chauhan, M. Vatsa, R. Singh, "A Robust Skin Color Based Face Detection Algorithm", Tamkang
Journal of Science and Engineering, 6(4), pp. 227–234 (2003)
[18] Shamna , Paul Augustine, Tripti C , “An Exploratory Survey on Various Face Recognition Methods Using
Component Analysis”, Vol. 2, Issue 5, May 2013.

680 B.Grace Jemimah, K.Rajendra Babu, K.Kamar Taz, C.Nagaraju

You might also like