Hits:

Indexed by:期刊论文

Date of Publication:2011-11-01

Journal:Advances in Information Sciences and Service Sciences

Included Journals:EI、Scopus

Volume:3

Issue:10

Page Number:345-354

ISSN No.:19763700

Abstract:Very recently, the sparse representation theory in pattern recognition has aroused widespread concern. Herein, we investigated the sparse representation-based face recognition methods. In order to make the representation coefficient vector sparser, a Gabor feature-based sparse representation classification (GSRC) method was presented, which used the holistic Gabor features to construct dictionary to enhance the robustness for the variations of illumination, expression, occlusion and pose. For the GSRC method, the local information provided by the spatial locations of Gabor features is not completely exploited. To rectify this, local Gabor features are first extracted from some spatially partitioned non-overlapping local patches of face images, and then used to construct sparse representation classifiers (component classifiers). Finally, all these component classifiers are combined to form an ensemble classifier by fuzzy fusion method for final classification. Experimental results on the ORL, AR and FERET face databases verified the feasibility and effectiveness of the proposed methods.