Abstract: A new kind of classifier based on singular value decomposition of face images which solves the small sample size problem in face recognition tasks is presented in this paper.The method of renewing classifier using the incorrectly classified face image to improve the performance of classifiers is also given.Experimental results show that high recognition rate can be obtained by using only one sample per person and that the renewal of classifiers using only one incorrectly classified sample can further improve the performance of classifiers.

Key words: singular value decomposition, classifier, feature subspace, face recognition