论文部分内容阅读
The human face recognition system becomes more and more universal in different parts of our society like governments, banks and social welfare. How to improve the efficiency of the discrimination of human face is the most significant issue in the human face recognition system. And it is also a hot issue in the high dimensional analysis.In this paper, we compared the performance of the facial recognition of three methods, EMPCA, Sparse PCA and Kernel PCA. These three methods based on the general idea of Principal Component Analysis and Fishers Linear Discriminant. Our study based on the ORL face data base. EMPCA used same idea of the classic PCA by using the EM algorithm. The Sparse PCA extends the classic PCA by adding sparsity constraint in order to explain the variables more clear. The Kernel PCA extends the classic PCA by using kernel methods. The comparison of these three methods in facial recognition can help us know much more about the applications of the dimension reduction methods.