Kernel direct discriminant analysis and its theoretical foundation

摘要

In this paper, the method of kernel direct discriminant analysis is analyzed from a new viewpoint and its theoretical foundation is revealed. Based on this result, an efficient and robust method is proposed. That is, the QR decomposition on the small-size matrix is adopted and then a small eigenvalue problem is solved. Finally, experimental results on ORL face database show that the proposed method is effective and feasible.