A GENERALIZED OPTIMAL SET OF DISCRIMINANT-VECTORS

A generalized optimal set of discriminant vectors for linear feature extraction is presented, First, the criteria of selecting the generalized optimal discriminant vectors are introduced, and then a unified solving method is derived to solve the vectors of the generalized optimal set in both cases of a large number of samples and a small number of samples. The experimental results show that the present method is superior to the Foley-Sammon method (Foley and Sammon, IEEE Trans. Comput. 24, 281-289 (1975)), the positive pseudoinverse method (Tian et al., Opt. Engng 25(7), 834-839 (1986)), the perturbation method (Hong and Yang, Pattern Recognition 24, 317-324 (1991)), and the matrix rank decomposition method (Cheng et al., Pattern Recognition 25, 101-111 (1992)) in terms of correct classification rate.