the most discrimination feature is the people's forehead,eyes, ears and the mouth. On the bases of this,to be four block in the face! So the face is divided into three small block based on line and into three small block based on the column! According to the face have symmetry properties,so features of matrix will be transformation! The new definition of the membership function used in maximum divergence difference for extracting the most identify characteristic vector,and then, according to the nearest neighbor classifier to classify the sample.