Kernel Heteroscedastic Discriminant Analysis for Face Recognition

Kernel method is a nonlinear feature extraction approach. Firstly, the samples in the original feature space are transformed into a higher dimensional feature space by nonlinear mapping. Then, linear approaches are used in the higher dimensional feature space, and thus nonlinear features of original samples are extracted. The Heteroscedastic Discriminant Analysis (HDA), in which the equal within-class scatters matrix constraint of Linear Discriminant Analysis (LDA) is removed and more discriminant information is achieved. In this paper, take the advantages of kernel method and HDA, kernel Heteroscedastic discriminant analysis (KHDA) is presented and used for face recognition. Experimental results based on Olivetti Research Laboratory (ORL), ORL and Yale mixture face database show the validity KHDA for face recognition.