A novel one-parameter regularized linear discriminant analysis for solving small sample size problem in face recognition

作者：Chen, WS^*; Yuen, PC; Huang, J; Dai, DQ

来源：ADVANCES IN BIOMETRIC PERSON AUTHENTICATION, PROCEEDINGS, SPRINGER-VERLAG BERLIN, HEIDELBERGER PLATZ 3, D-14197 BERLIN, GERMANY, 320-329, 2004.

In this paper, a new l-parameter regularized discriminant analysis (lPRDA) algorithm is developed to deal with the small sample size (S3) problem. The main limitation in regularization is that the computational complexity of determining the optimal parameters is very high. In view of this limitation, we derive a single parameter (iota) explicit expression formula for determining the 3 parameters. A simple and efficient method is proposed to determine the value of iota. The proposed lPRLDA method for face recognition has been evaluated with two public available databases, namely ORL and FERET databases. The average recognition accuracy of 50 runs for ORL and FERET database are 96.65% and 94.00% respectively. Comparing with existing LDA-based methods in solving the S3 problem, the proposed lPRLDA method gives the best performance.

• 出版日期2004
• 单位深圳大学; 香港浸会大学; 中山大学