Face recognition for biometric advantage of being a non-contact proc recognition algorithms has been proposed The face recognition system mainly consis feature extraction / reduction and classifi most popular tool, Principal Component used for feature extraction. For classificatio distance classifiers as well as Support Vecto with various kernels are used. Radial bas kernel in SVM is one of the widely use purpose. In this paper, Mexican hat wa SVM is proposed for classification and the kernel with other classification methods a proposed kernel performs better in terms vectors compared to RBF kernel and the also high with less number of features.