In this paper, an approach to face recognition is proposed, in which SVM combined with nearest center classification (NCC) is used as the classifier. The philosophy behind this is based on that idea that their discriminative capabilities are not totally overlapped so that NCC may work on the samples that SVMs fail. Firstly, the principal component analysis is used to reduce dimension and extract feature. Then support vector machine (one-to-other scheme) combined with nearest center classifier used for classification. We conduct the experiment on the base of ORL face database with our method and three other decision rules for their comparison. The experiment result is presented and discussed, which shows the effectiveness of the strategy described.
展开▼
