Performance Improvement in Classification Rate of Appearance Based Statistical Face Recognition Methods using SVM classifier

This paper compares the performance improvement in recognition rate of different face recognition methods. The face recognition methods such as 1dPCA, 2dPCA, KPCA, ICA and FDA usually use Euclidean distance and in some cases they use the cosine similarity function. Instead of traditional classification and distance measurement methods, SVM classifier is used for classification. The SVM classifier discussed here uses feature extracted from different face recognition methods. For testing the face recognition algorithms standard Yale face database is used. The experimental results show that the SVM classifier outperforms traditional classification and distance measurement methods. Further, this paper analysis the role of standard deviation parameter of RBF kernel on face recognition accuracy.