Sparse margin based discriminant analysis for face recognition

The existing margin-based discriminant analysis methods, which use K-nearest neighbor technique to characterize the margin, such as nonparametric discriminant analysis (NDA). These methods encounter a common problem, that is, the nearest neighbor parameter K must be chosen in advance. How to choose an optimal K is a theoretically difficult problem. In this paper, we present a new marginal characterization method using the sparse representation, which can successfully avoid the difficulty of the parameter selection. The effectiveness of the proposed method is evaluated through the experiments on AR and Extended Yale B database, and the experimental results show the fact that the performance of the proposed method superiors to the state-of-the-art feature extraction methods.