An improved face recognition algorithm based on SVD

This paper presents an improved face recognition algorithm based on SVD. The algorithm uses a sampling window to sample overlapping from top to bottom. According to all the matrices obtained by overlapping sampling, we count the singular values for all the matrices, reduce dimension and compression, and get k maximal singular values of every matrix that are formed a group of vectors. The group of vectors is acted as the features of the image to recognize. Experiments show that the recognition rate of the improved algorithm is higher.