Learning kernel subspace classifier for robust face recognition

Bailing Zhang, Sheng Uei Guan, Hanseok Ko

Research output: Contribution to journalArticle

Abstract

Subspace classifiers are very important in pattern recognition in which pattern classes are described in terms of linear subspaces spanned by their respective basis vectors. To overcome the limitations of linear methods, kernel based subspace models have been proposed in the past by applying the Kernel Principal Component Analysis (KPCA). However, the projection variance in the kernel space as applied in the previously proposed kernel subspace methods, is not a good criteria for the data representation and they simply fail in many recognition problems. We address this issue by proposing a learning kernel subspace classifier which attempts to reconstruct data in the input space through the kernel subspace projection. Comparing with the pre-image methods, we emphasize the problem of how to use a kernel subspace as a model to describe input space rather than finding an approximate pre-image for each input by minimization of the reconstruction error in the kernel space. Experimental results on occluded face recognition demonstrated the efficiency of the proposed method.

Original languageEnglish
Pages (from-to)103-108
Number of pages6
JournalInternational Journal of Soft Computing
Volume4
Issue number2
Publication statusPublished - 2009 Oct 5

Fingerprint

Face recognition
Face Recognition
Classifiers
Classifier
Subspace
kernel
Principal component analysis
Pattern recognition
Kernel Methods
Projection
Kernel Principal Component Analysis
Subspace Methods
Pattern Recognition
Learning
Experimental Results
Model

Keywords

  • Kernel method
  • Principal component analysis
  • Robust face recognition
  • Subspace classifier

ASJC Scopus subject areas

  • Software
  • Theoretical Computer Science
  • Modelling and Simulation

Cite this

Learning kernel subspace classifier for robust face recognition. / Zhang, Bailing; Guan, Sheng Uei; Ko, Hanseok.

In: International Journal of Soft Computing, Vol. 4, No. 2, 05.10.2009, p. 103-108.

Research output: Contribution to journalArticle

Zhang, Bailing ; Guan, Sheng Uei ; Ko, Hanseok. / Learning kernel subspace classifier for robust face recognition. In: International Journal of Soft Computing. 2009 ; Vol. 4, No. 2. pp. 103-108.
@article{065f19ea90f9423db65000e620e6fa1b,
title = "Learning kernel subspace classifier for robust face recognition",
abstract = "Subspace classifiers are very important in pattern recognition in which pattern classes are described in terms of linear subspaces spanned by their respective basis vectors. To overcome the limitations of linear methods, kernel based subspace models have been proposed in the past by applying the Kernel Principal Component Analysis (KPCA). However, the projection variance in the kernel space as applied in the previously proposed kernel subspace methods, is not a good criteria for the data representation and they simply fail in many recognition problems. We address this issue by proposing a learning kernel subspace classifier which attempts to reconstruct data in the input space through the kernel subspace projection. Comparing with the pre-image methods, we emphasize the problem of how to use a kernel subspace as a model to describe input space rather than finding an approximate pre-image for each input by minimization of the reconstruction error in the kernel space. Experimental results on occluded face recognition demonstrated the efficiency of the proposed method.",
keywords = "Kernel method, Principal component analysis, Robust face recognition, Subspace classifier",
author = "Bailing Zhang and Guan, {Sheng Uei} and Hanseok Ko",
year = "2009",
month = "10",
day = "5",
language = "English",
volume = "4",
pages = "103--108",
journal = "International Journal of Soft Computing",
issn = "1816-9503",
publisher = "Medwell Publishing",
number = "2",

}

TY - JOUR

T1 - Learning kernel subspace classifier for robust face recognition

AU - Zhang, Bailing

AU - Guan, Sheng Uei

AU - Ko, Hanseok

PY - 2009/10/5

Y1 - 2009/10/5

N2 - Subspace classifiers are very important in pattern recognition in which pattern classes are described in terms of linear subspaces spanned by their respective basis vectors. To overcome the limitations of linear methods, kernel based subspace models have been proposed in the past by applying the Kernel Principal Component Analysis (KPCA). However, the projection variance in the kernel space as applied in the previously proposed kernel subspace methods, is not a good criteria for the data representation and they simply fail in many recognition problems. We address this issue by proposing a learning kernel subspace classifier which attempts to reconstruct data in the input space through the kernel subspace projection. Comparing with the pre-image methods, we emphasize the problem of how to use a kernel subspace as a model to describe input space rather than finding an approximate pre-image for each input by minimization of the reconstruction error in the kernel space. Experimental results on occluded face recognition demonstrated the efficiency of the proposed method.

AB - Subspace classifiers are very important in pattern recognition in which pattern classes are described in terms of linear subspaces spanned by their respective basis vectors. To overcome the limitations of linear methods, kernel based subspace models have been proposed in the past by applying the Kernel Principal Component Analysis (KPCA). However, the projection variance in the kernel space as applied in the previously proposed kernel subspace methods, is not a good criteria for the data representation and they simply fail in many recognition problems. We address this issue by proposing a learning kernel subspace classifier which attempts to reconstruct data in the input space through the kernel subspace projection. Comparing with the pre-image methods, we emphasize the problem of how to use a kernel subspace as a model to describe input space rather than finding an approximate pre-image for each input by minimization of the reconstruction error in the kernel space. Experimental results on occluded face recognition demonstrated the efficiency of the proposed method.

KW - Kernel method

KW - Principal component analysis

KW - Robust face recognition

KW - Subspace classifier

UR - http://www.scopus.com/inward/record.url?scp=70349515548&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=70349515548&partnerID=8YFLogxK

M3 - Article

VL - 4

SP - 103

EP - 108

JO - International Journal of Soft Computing

JF - International Journal of Soft Computing

SN - 1816-9503

IS - 2

ER -