期刊名称：International Journal of Computer Science and Information Technologies
电子版ISSN：0975-9646
出版年度：2015
卷号：6
期号：4
页码：3897-3903
出版社：TechScience Publications
摘要：This work presents the extraction of entire Gabor features for efficient expression recognition and classification. Phase information available in Gabor filter bank is not properly utilized in several existing works for face and expression recognition. In this work both Gabor magnitude feature vector (GMFV) and Gabor phase congruency vectors (GPFV) are projected separately by subspace methods with respect to preserving non redundant data and reducing redundant coefficients. Locality preserving projection (LPP) subspace method is used for preserving and projecting the Gabor vector feature space. Projected vectors are normalized and fused. This EGLPP approach is tested with Yale and FD database respectively. Proposed approach improves the recognition rate while compared with EGPCA, EGICA and EGKPCA approaches. Support vector machine classifier is used for expression classification.
