Linear Discriminant Analysis for Human Face Recognition

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 04 Issue: 08 | Aug -2017 www.irjet.net p-ISSN: 2395-0072
© 2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 1005
Linear Discriminant Analysis for Human Face Recognition
Mr. Rahul M. Ohol1, Mrs. Shilpa R. Ohol2
1Lecturer, Dept. Of Medical Electronics Govt. Polytechnic Miraj, Maharashtra, India
2 Asst. Professor, Dept. Of Mathematics P.V.P.I.Technology, Sangli, Maharashtra, India
----------------------------------------------------------------------------------***------------------------------------------------------------------------------
Abstract – There are many possible techniques for
classification of data. PrincipalComponentAnalysis(PCA) and
Linear Discriminant Analysis (LDA) are two commonly used
techniques for data classification and dimensionality
reduction. LinearDiscriminantAnalysiseasilyhandlesthe case
where the within-class frequencies are unequal and their
performance has been examined on randomly generated test
data. This method maximizes the ratio of between-class
variance to the within-classvarianceinanyparticulardataset
thereby guaranteeing maximal separability. The useof Linear
Discriminant Analysis for data classification is applied to
classification problem in face recognition and speech
recognition.
Key Words: Principal Component Analysis(PCA),Linear
Discriminant Analysis (LDA), face recognition, data
classification
1. INTRODUCTION
Biometrics is an old Greek word, “Bio”, meaning life and
“Metric” the measure of, so Biometrics is in essence, the
measure of life. Simply defined, Biometrics is the automated
use of physiological or behavioural characteristics to
determine or verify identity of a person.A concise definition
of biometrics is “The automatic recognitionofa personusing
distinguishing traits.” A more expansive definition of
biometrics is “Any automatically measurable, robust and
distinctive physical characteristic or personal trait that can
be used to identify an individual or verify the claimed
identity of an individual.” This definition requires
elaboration. Biometrics is the science of verifying the
identity of an individual through physiological
measurements or behavioural traits. Since biometric
identifiers are associated permanently with the user and
they are more reliable than token or knowledge based
authentication methods. Biometrics offers several
advantages over traditional security measures [1].
1.1 Why Face Recognition?
Among the biometrics, the face is the most natural
physiological characteristic to recognize each other. Hence,
people consider face a “good” biometric for automatic
identity recognition systems.There are a number of reasons
to choose face recognition. These are as follows
 It is non-intrusive and requires no physical
interaction on behalf of the user.The system
captures faces of people in public areas, which
minimizes legal concerns for reasons explained
below. Moreover, since faces can be captured from
some distance away, facial recognition canbedone
without any physical contact.
 The acquisition process can be performed with a
limited person cooperation.
 It is accurate and allows for high enrolment and
verification rates.
 It does not require an expert to interpret the
comparisons.
 It can use the existing hardware infrastructure i.e.
existing cameras and image capture devices.
 It is the only biometric technology that allows you
to perform passive identification in a one-to-many
environment.
2 Linear Discriminant Analysis(LDA)
Originally developed in 1936 by R.A. Fisher, discriminant
analysis is a classic method of classification that has stood
the test of time. Discriminant analysisoftenproducesmodels
whose accuracyapproaches (andoccasionallyexceeds)more
complex modern methods.Discriminantanalysiscanbeused
only for classification (i.e., witha categorical targetvariable),
not for regression. The target variablemayhavetwoormore
categories. It is also known as Fisher Discriminant Analysis
(FDA).
2.1 Defination of groups
The groups to be discriminated can be defined either
naturally by the problem under investigation, or by some
preceding analysis, such as a cluster analysis. The number of
groups is not restricted to two, although the discrimination
between two groups isthemostcommonapproach.Notethat
the number of groups must not exceed the number of
variables describing the data set. Anotherprerequisiteisthat
the groups have the samecovariancestructure(i.e.theymust
be comparable).
Fig -1: Cluster analysis

2.2 Estimation of the parameters of the
discriminating function
There is only one direction of the discriminating line which
yields the best separation results. The determination of the
coefficients of the discriminating function is quite simple. In
principle, the discriminating function is formed in such a
way that the separation (=distance) between the groups is
maximized, and the distance within thegroupsisminimized.
A transformation function is found that maximizes the ratio
of between-class variance to within-class variance as
illustrated by this figure 4.4
Fig -2: Good class seperation
The transformation seeks to rotate the axes so that when
the categories are projected on the new axes, the differences
between the groups are maximized. The following figure
shows two rotates axes. Projection to the lower right axis
achieves the maximum separation between the categories;
projection to the lower left axis yields the worst separation.
Fig -3: Class feature data
2.3 Mathematical Operations
Fisher discriminates group images of the same class and
separates images of different classes. Images are projected
from N-dimensional space (where N is the number of pixels
in the image) to C-1 dimensional space (where C is the
number of classes of images). Forexample,considertwosets
of points in 2-dimensional space that are projected onto a
single line. Depending on the direction of the line, the points
can either be mixed together or separated. Fisher
discriminants find the line that best separates the points. To
identify a test image, the projected test imageiscompared to
each projected training image, and the test image is
identified as the closest training image.
As with eigenspace projection, training images are
projected into a subspace. The test images are projectedinto
the same subspace and identified usinga similaritymeasure.
What differs is how the subspace is calculated.Following are
the steps to follow to find the Fisher discriminants for a set
of images.
2.3.1 Calculate the within class scatter matrix
The within class scatter matrix measures the amount of
scatter between items in the same class. For the i thclass, a
scattermatrix (Si ) is calculated as the sum of the covariance
matrices of the centered images in that class.
T
xix
mixmixSi 
 ))((
mi is the mean of images in the class. The withinclassscatter
matrix (SW) is the sum of all scatter matrices.
i
c
i
W SS 

1
C is the number of classes.
2.3.2 Calculate the between class scatter matrix
The between class scatter matrix ( SB ) measurestheamount
of scatter between classes. It is calculated as the sum of the
covariance matrices ofthedifference betweenthetotal mean
and the mean of each class.


c
i
T
B mmimminiS
1
))((
where ni is the number of images in the class, mi is the
mean of the images in the class and m is the mean of all the
images.
2.3.3 Solve the generalized eigenvalue problem
Solve for the generalized eigenvectors (V ) and eigenvalues
() of the within class and between class scatter matrices.
VSVS WB 

2.3.4 Keep first C-l eigenvectors
Sort the eigenvectors by their associated eigenvalues from
high to low and keep the first C 1 eigenvectors. These
eigenvectors form the Fisher basis vectors.
2.3.5 Project images onto Fisher basis vectors
Project all the original (i.e. not centered)
images onto the Fisher basis vectors by calculating the dot
product of the image with each of the Fisher basis vectors.
The original images are projected onto this line because
these are the points that the line has been created to
discriminate, not the centered images.
Following are the steps to follow to find the Fisher
discriminants of a set of imagesbyfirst projectingtheimages
into any orthonormal basis.
1. Compute means Compute the mean of the imagesineach
class( mi )and the total mean of all images (m ).
2. Center the images in each class Subtract the mean of
each class from the images in that class.
3.Center the class means Subtract the total mean from the
class means.
4. Create a data matrix Combine the all images, side-by-
side, into one data matrix.
5.Find an orthonormal basis for thisdatamatrix Thiscan
be accomplished by using a QR Orthogonal-triangular
decomposition or by calculating the full set of eigenvectors
of the covariance matrix of the training data. Let the
orthonormal basis be U
6. Project all centered imagesintotheorthonormalbasis
Create vectors that are the dot product of the image and the
vectors in the orthonormal basis.
7. Project the centered means into the orthonormal
basis
8. Calculate the within class scatter matrix The within
class scatter matrix measures theamountofscatterbetween
items within the same class. For the ith class a scatter matrix
(Si ) is calculated as the sum of the covariance matrices of
the projected centered images for that class.
9. Calculate the between class scatter matrix The
between class scatter matrix (SB )
measures the amount of scatter between classes. It is
calculated as the sum of the covariance matrices of the
projected centered means of the classes, weighted by the
number of images in each class.
10. Solve the generalized eigenvalue problem
11. Keep the first C-l eigenvectors Sorttheeigenvectorsby
their associated eigenvalues from high to low and keep the
first C -1 eigenvectors. These are the Fisher basis vectors.
12. Project images onto eigenvectors: Project all the
rotated original (i.e. not centered)
images onto the Fisher basis vectors. First project the
original images into the orthonormal basis, and then project
these projected images onto the Fisher basis vectors. The
original rotated images are projected onto this line because
these are the points that the line has been created to
discriminate, not the centered images.
3. CONCLUSIONS
LDA attempt to maximize the between class scatter, while
minimizing the within class scatter. In other words, moves
images of the same class closer together, while moving
images of different classes furtherapart. Lightingconditions,
Image quality, Pose orientation plays an important role in
face recognition system
REFERENCES
[1] John D. Woodward, Jr., Christopher Horn, Julius Gatune
and Aryn Thomas, Biometrics A Look at Facial
Recognition, Prepared for the Virginia State Crime
Commission, Published by RAND,(2003).
[2]Gian Luca Marcialis and Fabio Roli, “Decision - level
fusion of PCA and LDA-based face recognition
algorithms,” International Journal of Image and Graphics,
Vol. 6, No.2, pp 293-311, (2006).
[3]Juwei Lu, Kostantinos N. Plataniotis, and Anastasios N.
Venetsanopoulos, “Face Recognition Using LDA-Based
Algorithms,” IEEE Transactions on Neural Networks,
Vol. 14, No.1, January 2003.
[4] A tutorial on Principal Component Analysis by Lindsay I
Smith February 26,2002

Linear Discriminant Analysis for Human Face Recognition

More Related Content

What's hot (19)

Similar to Linear Discriminant Analysis for Human Face Recognition (20)

More from IRJET Journal (20)

Recently uploaded (20)

Linear Discriminant Analysis for Human Face Recognition