A complete fuzzy discriminant analysis approach for face recognition

摘要: In this paper, some studies have been made on the essence of fuzzy linear discriminant analysis (F-LDA) algorithm and support vector machine (FSVM) classifier, respectively. As a kernel-based learning machine, FSVM is represented with membership function while realizing same classification results that conventional pair-wise classification. It outperforms other machines especially when unclassifiable regions still remain in those classifiers. However, serious drawback computation requirement increases rapidly increase number classes training sample size. To address problem, an improved method combines advantages decision tree, called DT-FSVM, proposed firstly. Furthermore, process feature extraction, reformative F-LDA based k-nearest neighbors (FKNN) implemented to achieve distribution information each original grade, which incorporated into redefinition scatter matrices. particular, considering fact outlier samples patterns may adverse influence result, we developed novel using relaxed normalized condition definition function. Thus, limitation from effectively alleviated. Finally, by making full use set theory, complete (CF-LDA) framework combining (RF-LDA) extraction DT-FSVM classifier. This hybrid applied face recognition extensive experimental conducted ORL NUST603 images databases demonstrate effectiveness algorithm.