PCA and FLDA are mainly used in face recognition and feature extraction. PCA uses eigen vector and FLDA uses within class scatter matrix and between class scatter matrix. When within class matrix becomes singular, it cannot be evaluated. A new method called semi-discrete decomposition is used in single image per person problems. The performance of this method is tested on 4-data bases, namely ORL, UMIST, Poly u-NIR, YALE. The proposed method performs better than SVD based approach and QRCP based approach in terms of recognition rate with training times in two times higher than QRCP.
展开▼