Face Recognition with Single Training Sample per Person Using Sparse Representation

摘要

It is a great challenge for face recognition with single training sample per person. In this paper, we try to propose a new algorithm based sparse representation to solve this problem. The algorithm takes the two-dimensional training samples as the training set directly rather than image vectors. So we can obtain the dictionary of sparse representation only using one sample. The proposed algorithm includes training process and classification process. In training process all the class's dictionaries have been trained using KSVD algorithm. In classification process, the test sample has been projected to every trained dictionary, and then computes the reconstruction residual. At last the test sample is classified to the one who can get the minimum reconstruction residual. Experimental results show that the proposed method is efficient and it can achieve higher recognition accuracy than many existing schemes.