Topological principal component analysis for face encoding and recognition

被引：9

作者：

Pujol, A

Vitrià, J

Lumbreras, F

Villanueva, JJ

机构：

[1] Univ Autonoma Barcelona, Comp Vis Lab, Bellaterra 08193, Cerdanyola Vall, Spain

[2] Univ Autonoma Barcelona, Dept Informat, Bellaterra 08193, Cerdanyola Vall, Spain

来源：

PATTERN RECOGNITION LETTERS | 2001年 / 22卷 / 6-7期

关键词：

generalization; principal component analysis; face recognition; topological covariance matrix; covariance estimation;

D O I：

10.1016/S0167-8655(01)00027-7

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

Principal component analysis (PCA)-like methods make use of an estimation of the covariances between sample variables. This estimation does not take into account their topological relationships. This paper proposes how to use these relationships in order to estimate the covariances in a more robust way. The new method topological principal component analysis (TPCA) is tested using both face encoding and recognition experiments showing how the generalization capabilities of PCA are improved. (C) 2001 Elsevier Science B.V. All rights reserved.

引用

页码：769 / 776

页数：8

共 17 条

[1]

[Anonymous], 2005, NEURAL NETWORKS PATT

[2]

[Anonymous], 1994, STAT NEURAL NETWORKS, DOI DOI 10.1007/978-3-642-79119-2_1

[3] Eigenfaces vs. Fisherfaces: Recognition using class specific linear projection [J].