SVM-based feature extraction for face recognition

Sang Ki Kim, Youn Jung Park, Kar Ann Toh, Sang Youn Lee

Research output: Contribution to journalArticle

48 Citations (Scopus)

Abstract

The primary goal of linear discriminant analysis (LDA) in face feature extraction is to find an effective subspace for identity discrimination. The introduction of kernel trick has extended the LDA to nonlinear decision hypersurface. However, there remained inherent limitations for the nonlinear LDA to deal with physical applications under complex environmental factors. These limitations include the use of a common covariance function among each class, and the limited dimensionality inherent to the definition of the between-class scatter. Since these problems are inherently caused by the definition of the Fisher's criterion itself, they may not be solvable under the conventional LDA framework. This paper proposes to adopt a margin-based between-class scatter and a regularization process to resolve the issue. Essentially, we redesign the between-class scatter matrix based on the SVM margins to facilitate an effective and reliable feature extraction. This is followed by a regularization of the within-class scatter matrix. Extensive empirical experiments are performed to compare the proposed method with several other variants of the LDA method using the FERET, AR, and CMU-PIE databases.

Original languageEnglish
Pages (from-to)2871-2881
Number of pages11
JournalPattern Recognition
Volume43
Issue number8
DOIs
Publication statusPublished - 2010 Aug 1

Fingerprint

Discriminant analysis
Face recognition
Feature extraction
Experiments

All Science Journal Classification (ASJC) codes

  • Software
  • Signal Processing
  • Computer Vision and Pattern Recognition
  • Artificial Intelligence

Cite this

Kim, Sang Ki ; Park, Youn Jung ; Toh, Kar Ann ; Lee, Sang Youn. / SVM-based feature extraction for face recognition. In: Pattern Recognition. 2010 ; Vol. 43, No. 8. pp. 2871-2881.
@article{99083c196a9a447c85cdd995ebfa3e0f,
title = "SVM-based feature extraction for face recognition",
abstract = "The primary goal of linear discriminant analysis (LDA) in face feature extraction is to find an effective subspace for identity discrimination. The introduction of kernel trick has extended the LDA to nonlinear decision hypersurface. However, there remained inherent limitations for the nonlinear LDA to deal with physical applications under complex environmental factors. These limitations include the use of a common covariance function among each class, and the limited dimensionality inherent to the definition of the between-class scatter. Since these problems are inherently caused by the definition of the Fisher's criterion itself, they may not be solvable under the conventional LDA framework. This paper proposes to adopt a margin-based between-class scatter and a regularization process to resolve the issue. Essentially, we redesign the between-class scatter matrix based on the SVM margins to facilitate an effective and reliable feature extraction. This is followed by a regularization of the within-class scatter matrix. Extensive empirical experiments are performed to compare the proposed method with several other variants of the LDA method using the FERET, AR, and CMU-PIE databases.",
author = "Kim, {Sang Ki} and Park, {Youn Jung} and Toh, {Kar Ann} and Lee, {Sang Youn}",
year = "2010",
month = "8",
day = "1",
doi = "10.1016/j.patcog.2010.03.008",
language = "English",
volume = "43",
pages = "2871--2881",
journal = "Pattern Recognition",
issn = "0031-3203",
publisher = "Elsevier Limited",
number = "8",

}

SVM-based feature extraction for face recognition. / Kim, Sang Ki; Park, Youn Jung; Toh, Kar Ann; Lee, Sang Youn.

In: Pattern Recognition, Vol. 43, No. 8, 01.08.2010, p. 2871-2881.

Research output: Contribution to journalArticle

TY - JOUR

T1 - SVM-based feature extraction for face recognition

AU - Kim, Sang Ki

AU - Park, Youn Jung

AU - Toh, Kar Ann

AU - Lee, Sang Youn

PY - 2010/8/1

Y1 - 2010/8/1

N2 - The primary goal of linear discriminant analysis (LDA) in face feature extraction is to find an effective subspace for identity discrimination. The introduction of kernel trick has extended the LDA to nonlinear decision hypersurface. However, there remained inherent limitations for the nonlinear LDA to deal with physical applications under complex environmental factors. These limitations include the use of a common covariance function among each class, and the limited dimensionality inherent to the definition of the between-class scatter. Since these problems are inherently caused by the definition of the Fisher's criterion itself, they may not be solvable under the conventional LDA framework. This paper proposes to adopt a margin-based between-class scatter and a regularization process to resolve the issue. Essentially, we redesign the between-class scatter matrix based on the SVM margins to facilitate an effective and reliable feature extraction. This is followed by a regularization of the within-class scatter matrix. Extensive empirical experiments are performed to compare the proposed method with several other variants of the LDA method using the FERET, AR, and CMU-PIE databases.

AB - The primary goal of linear discriminant analysis (LDA) in face feature extraction is to find an effective subspace for identity discrimination. The introduction of kernel trick has extended the LDA to nonlinear decision hypersurface. However, there remained inherent limitations for the nonlinear LDA to deal with physical applications under complex environmental factors. These limitations include the use of a common covariance function among each class, and the limited dimensionality inherent to the definition of the between-class scatter. Since these problems are inherently caused by the definition of the Fisher's criterion itself, they may not be solvable under the conventional LDA framework. This paper proposes to adopt a margin-based between-class scatter and a regularization process to resolve the issue. Essentially, we redesign the between-class scatter matrix based on the SVM margins to facilitate an effective and reliable feature extraction. This is followed by a regularization of the within-class scatter matrix. Extensive empirical experiments are performed to compare the proposed method with several other variants of the LDA method using the FERET, AR, and CMU-PIE databases.

UR - http://www.scopus.com/inward/record.url?scp=77951258869&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77951258869&partnerID=8YFLogxK

U2 - 10.1016/j.patcog.2010.03.008

DO - 10.1016/j.patcog.2010.03.008

M3 - Article

VL - 43

SP - 2871

EP - 2881

JO - Pattern Recognition

JF - Pattern Recognition

SN - 0031-3203

IS - 8

ER -