Abstract:

We propose a SRC optimal discriminant projection method according to the classification criteria of the sparse representation classifier (SRC). There are two optimized goals: one is the sparse reconstruction residual betweenclass and withinclass of the data set, the other is the distinction of samples in the data set, which can achieve a higher classification accuracy of the SRC when the samples are projected into a low dimension space. Face recognition experiments conducted on the AR and Yale databases show that the proposed method is more effective and robust in comparison with several other popular methods.

Key words: sparse representation classifier (SRC), discriminant projection, face recognition, classification performance