Face recognition method based on Within-class Clustering SVM

A face recognition method based on within-class Clustering SVM (CCSVM) is presented in this paper in order to decrease the negative effect caused by noisy training samples in the recognition process. Based on the discontinuity of finite samples distribution in the high dimension space and the existence of noisy samples, first, we re-cluster samples within the class, find out the cluster centres to form the virtual classes, and then divide virtual classes of all the classes by SVM. Experiment results show that this method follows the distribution law of points in high-dimensional space and can achieve better performance than some traditional methods.