IRJET-Analysis of Face Recognition System for Different Classifier
2 likes•170 views
The document presents an analysis of a face recognition system utilizing different classifiers, specifically focusing on Kernel Fisher Analysis for feature extraction and evaluating the performance of Euclidean distance and Support Vector Machine classifiers. It details preprocessing steps, feature extraction methods, and experimental results from testing on a dataset of real-time images. The findings indicate that the SVM classifier outperforms the Euclidean classifier with a higher recognition rate of 58% compared to 35%.
![International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056
Volume: 02 Issue: 01 | Apr-2015 www.irjet.net p-ISSN: 2395-0072
© 2015, IRJET.NET- All Rights Reserved Page 404
ANALYSIS OF FACE RECOGNITION SYSTEM FOR DIFFERENT
CLASSIFIER
M.MANIMOZHI1, A. JOHN DHANASEELY2
1 M.E Applied Electronics, IFET COLLEGE ENGINEERING, Villupuram
2 Associate Professor, IFET COLLEGE OF ENGINEERING, Villupuram
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Face recognition plays vital role for
authenticating system. Human Face recognition is a
challenging task in computer vision and pattern
recognition. Face recognition has attracted much
attention due to its potential value in security and law
enforcement applications and its theoretical
challenges. Different methods are used for feature
extraction and classification. Kernel fisher analysis is
used for feature extraction. The performance analysis
for Euclidean, support vector machine is evaluated.
The whole process is done using MATLAB software. A
set of 10 person real time images is taken for our
work. The classifier recognizes the similar posture as
an output.
Key Words: Euclidean distance and SVM classifier,
Kernel Fisher Analysis.
1.INTRODUCTION
Over the past few years, the user authentication is
increasingly important because the security control is
required everywhere. Traditionally, ID cards and
passwords are popular for authentication.
Image processing refers to the processing of an image in
digital form. Modern video cameras may directly take the
image in digital form but generally images are originated
in optical form. They are captured by video cameras and
digitalized. The digitalization process includes sampling,
quantization.
Applications of face recognition are widespread.
Perhaps the most obvious is that of human computer
interaction. One could make computers easier to use if
when one simply sat down at a computer terminal, the
computer could identify the user by name and
automatically load personal preferences. This
identification could even be useful in enhancing other
technologies such as speech recognition, since if the
computer can identify the individual who is speaking the
voice patterns being observed can be accurately classified
against the known individual’s voice.
An image may be defined as a two-dimensional function,
f(x, y), where x and y are spatial (plane) coordinates, and
the amplitude of at any pair of coordinates (x, y) is called
the intensity or gray level of the image at that point.
Kfa(Kernel Fisher Analysis) is proposed for feature
extraction. The main idea of this method is to yield a
nonlinear discriminate analysis in the higher space.Two
different types of classifier are used and performance
analysis of both the classifier is evaluated.
The paper is organized as follows. Section II illustrates
the related work. Section III employs the proposed work
and describes about classifier Section IV describes the
experimental results. Section V will provide the
performance analysis. Section VI will provide the
conclusion and future work.
2.RELATED WORK
In [1] Heterogeneous Face Recognition (HFR)
framework which includes prototype random subspace is
used . A set of training subjects with an image from each
modality can be used as a prototype and depending on the
modality of a new image, the image from each prototype
subject can be selected from the corresponding modality.
Random sampling provides better result. The accuracy of
each of the tested HFR is low[1].
In [2] Locality constraints kernel space is used. It
includes coupled spectral regression and coupled
discriminate analysis. It is an effective subspace learning
framework, deducing from the graph embedding view of
subspace learning. Difference in appearance from different
modalities is reduced. Local structure of the data is not
preserved [2].
In [3] Simplified Weber level descriptor is used. It
extracts the geometry of a face and distribution of
geometric features in an image. It finds the most of](https://image.slidesharecdn.com/irjet-v2i173-150511141534-lva1-app6891/85/IRJET-Analysis-of-Face-Recognition-System-for-Different-Classifier-1-320.jpg)
![International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056
Volume: 02 Issue: 01 | Jan-2015 www.irjet.net p-ISSN: 2395-0072
© 2015, IRJET.NET- All Rights Reserved Page 405
effective low dimensional embedding for the original
sample data. Two database results are employed. Accuracy
is low [3].
3. PROPOSED WORK
Real time images are given as input to the pre-
processing method.Preprocessing is the basic step in an
image processing. It is used to remove the unwanted noise
that is present in the image. It is used to reduce the
complexity for further process. The input image is
converted into grey scale image. The histogram of an
image is calculated .It is the graphical representation of
grey level values in the x-axis and no of pixels in the y axis.
Smoothing is the process of reducing sharp transitions .It
is also used to reduce noise and blurring that are present
in the image.
The main objective of feature extraction is to find
meaningful low-dimensional representations of high
dimensional patterns such that the inherent data
structures and relations are revealed. Feature extraction is
an important technique for high-dimensional data, which
can avoid the curse of dimensionality and improve
computational efficiency in pattern classification.
Kernel Fisher Analysis (KFA) is proposed for feature
extraction. The aim of KFA is to achieve a nonlinear
discriminate analysis in the higher space. In this project
local features such as eye, eyebrow, nose and mouth are
extracted.
. The classifier used for training and testing is
Euclidean and Support Vector machine (SVM) classifier.
Fig-1: Block Diagram Of Face Recognition
In this paper, real time images are taken and a database
is created. The database contains 10 person images in
which each person having 10 different postures. For each
person 2 images are trained and 8 images are used for
testing.
The Euclidian distance classifier is one of the
classification techniques in area where clusters of points
representing the different entities to be classified are
spaced far apart in feature space. The idea behind the
Euclidian distance classifier is that one computes the
average of several training vectors for each possible
categorization and then classifies a given test vector by
determining to which cluster the average is the vector
nearest.
SVM (Support Vector Machine) constructs the hyper
plane that is used for classification .In this positive support
vectors and negative support vectors are used for posture
classification.
4. RESULT
The real time image is given input to the pre-
processing method. Fig 2 shows the image chosen for face
recognition.
Fig-2: Input selection
Fig 3 shows the grayscale image.RGB image is converted
into grayscale for better result.
Fig-3: Grayscale image
Input Image
Preprocessing
Feature Extraction
(Kernel Fisher Analysis)
Classifier (Euclidean or
Support Vector Machine)](https://image.slidesharecdn.com/irjet-v2i173-150511141534-lva1-app6891/85/IRJET-Analysis-of-Face-Recognition-System-for-Different-Classifier-2-320.jpg)
![International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056
Volume: 02 Issue: 01 | Jan-2015 www.irjet.net p-ISSN: 2395-0072
© 2015, IRJET.NET- All Rights Reserved Page 407
5. PERFORMACE ANALYSIS
The overall recognition rate is evaluated for both the
classifier. In this project Total number of real time person
we took is 10.For each person 10 different posture is
taken. For training 2 images and for testing 8 images is
taken for each person.
RECOGNITION RATE = No. of. Recognized images/
No. of. Testing images
Table-1: Recognition rate percentage
CLASSIFIER RECOGNITION RATE
PERCENTAGE
EUCLIDEAN DISTANCE 35
SVM 58
6.CONCLUSION AND FUTURE WORK
Face recognition analysis for different classifier is
evaluated. The real time images are preprocessed and
feature is extracted using kernel fisher analysis algorithm.
For classification, two different classifier is used.SVM
classifier provides better result (i.e.) 23% higher than
Euclidean distance classifier.
The future work includes using of alternate algorithm
for feature extraction and using of various databases. The
testing can be done by many classifiers for recognizing
purposes.
REFERENCES
[1] B. Klare and A. K. Jain, “Heterogeneous face
recognition using kernel prototype similarities,” IEEE
Trans. Pattern Anal. Mach. Intel., vol. 35,no. 6, pp. 1410–
1422, Jun. 2013.
[2] Z. Lei, S. Liao, A. K. Jain, and S. Z. Li, “Coupled
discriminant analysis for heterogeneous face recognition,”
IEEE Trans. Inf. Forensics Security,vol. 7, no. 6, pp. 1707–
1716, Dec. 2012.
[3] S. Liao, D. Yi, Z. Lei, R. Qin, and S. Li, “Heterogeneous
face recognition from local structures of normalized
appearance,” in Proc. 3rd ICB, ,pp. 209–218,2009
[4] J. Yang, A. F. Franji, J. Y. Yang, D. Zhang, and Z. Jin,
“KPCA plus LDA:A complete kernel fisher discriminant
framework for feature extraction and recognition,” IEEE
Trans. Pattern Anal. Mach. Intell., vol. 27, no. 2, pp. 230–
244, Feb. 2005.
[5] K. M. Lam and H. Yan, "Locating and extracting the eye
in human face images", Pattern Recognition, Vol. 29, No.5
771-779, 1996.
[6] L.Sirovich and M. Kirby, "Low-dimensional procedure
for the characterization of human faces," J. Opt. Soc. Am. A,
Vol. 4, No. 3, 519-524, 1987.
[7] R. Brunelli and T. Poggio, ―Face recognition: Features
versus Templates‖, IEEE Trans. Pattern Analysis and
Machine Intelligence, 15(10): 1042-1052, 1993.
[8] Ergun Gumus, Niyazi Kilic, Ahmet Sertbas and Osman
N. Ucan(2011) “Evaluation of face recognition techniques
using PCA,wavelets and SVM” Expert Systems with
Applications 37 (2010)6404–6408
[9] M.Sathish Kumar, S.Sangeetha(2014) “Blur and
Illumination robust face recognition using SVM classifier”
INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH
IN ELECTRICAL, ELECTRONICS, INSTRUMENTATION AND
CONTROL ENGINEERING Vol. 2, Issue 1, January 2014
[10]Pedro Tome, Julian Fierrez, Ruben Vera-Rodriguez,
and Mark S. Nixon,” Soft Biometrics and Their Application
in Person Recognition at a Distance”, IEEE Transactions
On Information Forensics And Security, Vol. 9, No. 3,
March 2014.
[11] M.Manimozhi, S.Sangeetha,” kernel fisher analysis
based feature extraction for face recognition using
Euclidean classifier ”,IJIRCCE, vol 2,issue 11,Nov 2014.
BIOGRAPHIES
M.MANIMOZHI completed B.E(ECE) from
Srinivasan Engineering College in the year 2013 .Currently
pursuing M.E(APPLIED ELECTRONICS) fron IFET College
Of Engineering,villupuram.
A. John Dhanaseely completed B.E(ECE) from Anna
University, Guindy, Chennai in the year 1994 .She obtained
her M.E. degree in Applied Electronics in the year 2004.
She is currently pursuing her Ph.D degree in the area of AI
Techniques.](https://image.slidesharecdn.com/irjet-v2i173-150511141534-lva1-app6891/85/IRJET-Analysis-of-Face-Recognition-System-for-Different-Classifier-4-320.jpg)
IRJET-Analysis of Face Recognition System for Different Classifier
- 1. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056 Volume: 02 Issue: 01 | Apr-2015 www.irjet.net p-ISSN: 2395-0072 © 2015, IRJET.NET- All Rights Reserved Page 404 ANALYSIS OF FACE RECOGNITION SYSTEM FOR DIFFERENT CLASSIFIER M.MANIMOZHI1, A. JOHN DHANASEELY2 1 M.E Applied Electronics, IFET COLLEGE ENGINEERING, Villupuram 2 Associate Professor, IFET COLLEGE OF ENGINEERING, Villupuram ---------------------------------------------------------------------***--------------------------------------------------------------------- Abstract - Face recognition plays vital role for authenticating system. Human Face recognition is a challenging task in computer vision and pattern recognition. Face recognition has attracted much attention due to its potential value in security and law enforcement applications and its theoretical challenges. Different methods are used for feature extraction and classification. Kernel fisher analysis is used for feature extraction. The performance analysis for Euclidean, support vector machine is evaluated. The whole process is done using MATLAB software. A set of 10 person real time images is taken for our work. The classifier recognizes the similar posture as an output. Key Words: Euclidean distance and SVM classifier, Kernel Fisher Analysis. 1.INTRODUCTION Over the past few years, the user authentication is increasingly important because the security control is required everywhere. Traditionally, ID cards and passwords are popular for authentication. Image processing refers to the processing of an image in digital form. Modern video cameras may directly take the image in digital form but generally images are originated in optical form. They are captured by video cameras and digitalized. The digitalization process includes sampling, quantization. Applications of face recognition are widespread. Perhaps the most obvious is that of human computer interaction. One could make computers easier to use if when one simply sat down at a computer terminal, the computer could identify the user by name and automatically load personal preferences. This identification could even be useful in enhancing other technologies such as speech recognition, since if the computer can identify the individual who is speaking the voice patterns being observed can be accurately classified against the known individual’s voice. An image may be defined as a two-dimensional function, f(x, y), where x and y are spatial (plane) coordinates, and the amplitude of at any pair of coordinates (x, y) is called the intensity or gray level of the image at that point. Kfa(Kernel Fisher Analysis) is proposed for feature extraction. The main idea of this method is to yield a nonlinear discriminate analysis in the higher space.Two different types of classifier are used and performance analysis of both the classifier is evaluated. The paper is organized as follows. Section II illustrates the related work. Section III employs the proposed work and describes about classifier Section IV describes the experimental results. Section V will provide the performance analysis. Section VI will provide the conclusion and future work. 2.RELATED WORK In [1] Heterogeneous Face Recognition (HFR) framework which includes prototype random subspace is used . A set of training subjects with an image from each modality can be used as a prototype and depending on the modality of a new image, the image from each prototype subject can be selected from the corresponding modality. Random sampling provides better result. The accuracy of each of the tested HFR is low[1]. In [2] Locality constraints kernel space is used. It includes coupled spectral regression and coupled discriminate analysis. It is an effective subspace learning framework, deducing from the graph embedding view of subspace learning. Difference in appearance from different modalities is reduced. Local structure of the data is not preserved [2]. In [3] Simplified Weber level descriptor is used. It extracts the geometry of a face and distribution of geometric features in an image. It finds the most of
- 2. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056 Volume: 02 Issue: 01 | Jan-2015 www.irjet.net p-ISSN: 2395-0072 © 2015, IRJET.NET- All Rights Reserved Page 405 effective low dimensional embedding for the original sample data. Two database results are employed. Accuracy is low [3]. 3. PROPOSED WORK Real time images are given as input to the pre- processing method.Preprocessing is the basic step in an image processing. It is used to remove the unwanted noise that is present in the image. It is used to reduce the complexity for further process. The input image is converted into grey scale image. The histogram of an image is calculated .It is the graphical representation of grey level values in the x-axis and no of pixels in the y axis. Smoothing is the process of reducing sharp transitions .It is also used to reduce noise and blurring that are present in the image. The main objective of feature extraction is to find meaningful low-dimensional representations of high dimensional patterns such that the inherent data structures and relations are revealed. Feature extraction is an important technique for high-dimensional data, which can avoid the curse of dimensionality and improve computational efficiency in pattern classification. Kernel Fisher Analysis (KFA) is proposed for feature extraction. The aim of KFA is to achieve a nonlinear discriminate analysis in the higher space. In this project local features such as eye, eyebrow, nose and mouth are extracted. . The classifier used for training and testing is Euclidean and Support Vector machine (SVM) classifier. Fig-1: Block Diagram Of Face Recognition In this paper, real time images are taken and a database is created. The database contains 10 person images in which each person having 10 different postures. For each person 2 images are trained and 8 images are used for testing. The Euclidian distance classifier is one of the classification techniques in area where clusters of points representing the different entities to be classified are spaced far apart in feature space. The idea behind the Euclidian distance classifier is that one computes the average of several training vectors for each possible categorization and then classifies a given test vector by determining to which cluster the average is the vector nearest. SVM (Support Vector Machine) constructs the hyper plane that is used for classification .In this positive support vectors and negative support vectors are used for posture classification. 4. RESULT The real time image is given input to the pre- processing method. Fig 2 shows the image chosen for face recognition. Fig-2: Input selection Fig 3 shows the grayscale image.RGB image is converted into grayscale for better result. Fig-3: Grayscale image Input Image Preprocessing Feature Extraction (Kernel Fisher Analysis) Classifier (Euclidean or Support Vector Machine)
- 3. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056 Volume: 02 Issue: 01 | Jan-2015 www.irjet.net p-ISSN: 2395-0072 © 2015, IRJET.NET- All Rights Reserved Page 406 Fig-4: Filter image The above figure shows the filtered image. Filtering is to improve the image by filtering the unwanted noise. Local features are extracted in this project using kernel fisher analysis. The extracted feature is shown in fig 5 Fig-5: Feature Extraction Left eyebrow and Right eyebrow are separately extracted and an eye pair image from real time image. The face image is extracted at first with a fixed size. In this paper 100 real time images feature extraction is done. In this paper all the local features are extracted, due to this the classifier will analyzed the face efficiently. Fig-6: Face recognition using Euclidean classifier The above figure shows the result obtained by using Euclidean classifier. One of the trained images is given as input and the above posture is obtained for single person input. Similarly for all the 10 persons, output is obtained by giving trained image as an input. Fig-7: Face recognition using SVM classifier Fig 7 shows the result obtained using SVM classifier. The trained image is given as an input and the classifier checks the database and provides the similar posture to the input.
- 4. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056 Volume: 02 Issue: 01 | Jan-2015 www.irjet.net p-ISSN: 2395-0072 © 2015, IRJET.NET- All Rights Reserved Page 407 5. PERFORMACE ANALYSIS The overall recognition rate is evaluated for both the classifier. In this project Total number of real time person we took is 10.For each person 10 different posture is taken. For training 2 images and for testing 8 images is taken for each person. RECOGNITION RATE = No. of. Recognized images/ No. of. Testing images Table-1: Recognition rate percentage CLASSIFIER RECOGNITION RATE PERCENTAGE EUCLIDEAN DISTANCE 35 SVM 58 6.CONCLUSION AND FUTURE WORK Face recognition analysis for different classifier is evaluated. The real time images are preprocessed and feature is extracted using kernel fisher analysis algorithm. For classification, two different classifier is used.SVM classifier provides better result (i.e.) 23% higher than Euclidean distance classifier. The future work includes using of alternate algorithm for feature extraction and using of various databases. The testing can be done by many classifiers for recognizing purposes. REFERENCES [1] B. Klare and A. K. Jain, “Heterogeneous face recognition using kernel prototype similarities,” IEEE Trans. Pattern Anal. Mach. Intel., vol. 35,no. 6, pp. 1410– 1422, Jun. 2013. [2] Z. Lei, S. Liao, A. K. Jain, and S. Z. Li, “Coupled discriminant analysis for heterogeneous face recognition,” IEEE Trans. Inf. Forensics Security,vol. 7, no. 6, pp. 1707– 1716, Dec. 2012. [3] S. Liao, D. Yi, Z. Lei, R. Qin, and S. Li, “Heterogeneous face recognition from local structures of normalized appearance,” in Proc. 3rd ICB, ,pp. 209–218,2009 [4] J. Yang, A. F. Franji, J. Y. Yang, D. Zhang, and Z. Jin, “KPCA plus LDA:A complete kernel fisher discriminant framework for feature extraction and recognition,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 27, no. 2, pp. 230– 244, Feb. 2005. [5] K. M. Lam and H. Yan, "Locating and extracting the eye in human face images", Pattern Recognition, Vol. 29, No.5 771-779, 1996. [6] L.Sirovich and M. Kirby, "Low-dimensional procedure for the characterization of human faces," J. Opt. Soc. Am. A, Vol. 4, No. 3, 519-524, 1987. [7] R. Brunelli and T. Poggio, ―Face recognition: Features versus Templates‖, IEEE Trans. Pattern Analysis and Machine Intelligence, 15(10): 1042-1052, 1993. [8] Ergun Gumus, Niyazi Kilic, Ahmet Sertbas and Osman N. Ucan(2011) “Evaluation of face recognition techniques using PCA,wavelets and SVM” Expert Systems with Applications 37 (2010)6404–6408 [9] M.Sathish Kumar, S.Sangeetha(2014) “Blur and Illumination robust face recognition using SVM classifier” INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN ELECTRICAL, ELECTRONICS, INSTRUMENTATION AND CONTROL ENGINEERING Vol. 2, Issue 1, January 2014 [10]Pedro Tome, Julian Fierrez, Ruben Vera-Rodriguez, and Mark S. Nixon,” Soft Biometrics and Their Application in Person Recognition at a Distance”, IEEE Transactions On Information Forensics And Security, Vol. 9, No. 3, March 2014. [11] M.Manimozhi, S.Sangeetha,” kernel fisher analysis based feature extraction for face recognition using Euclidean classifier ”,IJIRCCE, vol 2,issue 11,Nov 2014. BIOGRAPHIES M.MANIMOZHI completed B.E(ECE) from Srinivasan Engineering College in the year 2013 .Currently pursuing M.E(APPLIED ELECTRONICS) fron IFET College Of Engineering,villupuram. A. John Dhanaseely completed B.E(ECE) from Anna University, Guindy, Chennai in the year 1994 .She obtained her M.E. degree in Applied Electronics in the year 2004. She is currently pursuing her Ph.D degree in the area of AI Techniques.