ICA Based Minimum Discriminant Analysis and Its Application to Face Recognition

Abstract

Face recognition is a very active field for research in the field of pattern recognition. To improve the performance of feature extraction in face recognition, a novel feature extraction method named as minimal linear discriminant analysis based on independent component analysis (ICA) is proposed. Therefore, the singular problem of the within-class scatter matrix will be avoided, and linear discriminant vectors with most discriminant information can be obtained. Experimental results on Yale and ORL face databases demonstrate that the recognition rate of the proposed method is more effective than that of the classical methods.