Sparse representation with nearest subspaces for face recognition

This paper proposes a supervised sparse representation based classification method for face recognition. The proposed method consists of two phases. The first phase seeks k nearest subspaces for the test sample from the c classes, and converts the c -class classification problem into a k -class problem, where c is the number of classes and k is a parameter. We can do this because, as validated by the experiments, the test sample belongs to one of these k subspaces with a very high probability. The second phase represents the test sample using these k nearest subspaces. This phase enhances the sparability by increasing the contribution of right subspace and suppressing those of the others. Via the above two phases, the proposed method introduces an approach to gain the sparsity in representation. The experimental results show the feasibility of the proposed method in face recognition.