Research On Face Recognition Algorithm Based On Sparse Representation

In recent years, biometric identification technology has become a kind of convenient, safe and reliable security technology because of the advantages of biological characteristics, which can’t be forgotten and lost. Face recognition technology applying computer technology, utilizing human face biological characteristics to distinguish the human faces, is recognized as the most "noninvasive", the most convenient and very "human" technology. It has become one of the hot topics in current computer technology research.Face recognition mainly includes two steps: facial feature extraction and classification. This paper first systematically introduces three classical algorithms produced in the development of the facial feature extraction: linear discriminate analysis(LDA), principal component analysis(PCA) and locality preserving projection(LPP). These three algorithms belong to the classical dimensionality reduction algorithm in the field of pattern recognition. They have greatly promoted the development of face recognition technology and been learned, emulated and compared so far. At present, more and more people pay attention to the face recognition based on sparse description. It is a new perspective to treat and deal with the problem of face recognition. The idea can be understood as: the face test samples can be approximated by the training samples, and then the test samples belong to the largest category in the expression. In this paper, I propose three improved conventional transformation methods(SRPCA, SRLDA and SRLPP) based on sparse description, and analyze the rationality of the improved conventional transformation method.There are many methods of face classification recognition. the method of sparse description developed in recent years is a new face recognition method, which has a completely different methodology. In this paper, we introduce two sparse representation classifiers: sparse representation classifier(SRC), weighted sparse representation classifier(WSRC). And on this basis we propose an improved weighted sparse representation classifier(IWSRC). Finally, the four algorithms(SRPCA, SRLDA, SRLPP and IWSRC) proposed in this paper are tested on four face databases(ORL, Yale, Yale B, AR), and the experimental results are analyzed and compared. Experiments show that their performance is much better than that of the traditional feature extraction method and sparse representation classifier.