On improvement of multiple discriminant analysis method for discriminative feature extraction

Two new algorithms-modified multiple discriminant analysis (MMDA) and difference principle component analysis (DPCA)-are presented and derived. The proposed algorithms are especially useful in automatic feature extraction for extremely similar patterns in a small category set. Experimental results for recognition of Chinese character fonts and handwritten numerals using MMDA and DPCA are presented. Compared with the traditional algorithms, MMDA and DPCA avoid the computational stability problem inherent in multiple discriminant analysis (MDA), and provide more effective feature metrics for pattern discrimination