The Problem Existed in Face Recognition Using SVD and Its Solution
SVD用于人脸识别存在的问题及解决方法

The singular value vector of face image is analyzed and its nature is revealed,i.e.,the singular values of face image are obtained by decomposing the face in the special basis space that is determined by the face image.Based on this result,the intrinsic reasons that singular value vector-based recognition methods have the low recognition accauracy are follows: singular value vectors of arbitrary two face images have not the equivalent basis space in general;there is no one-to-one correspondence between singular value vector and face image;singular value vectors are not separated.Finally,a novel face recognition method coined class estimated basis space method is presented.ORL,ORL-NWPU1 and ORL-NWPU2 database are used to test,and the experimental results show the correctness of analysis and the proposed method.