Face Recognition Using Weighted Pairwise Fuzzy Fisherface Method

A weighted pairwise fuzzy fisherface (WPF-fisherface) method is presented in this paper for face recognition. Fuzzy fisherface (F-fisherface) method is an extension of the fisherface method by incorporating the class membership to each training sample in calculating the scatter matrices. In this paper, we reformulate the fuzzy between-class scatter matrix in a pairwise form and then impose a weight on each pair of classes, where the weight aims to emphasize the different roles of the individual class pairs in the discrimination. By using the weighted pairwise form of the between-class scatter matrix, we can make use of the advantage of the class membership. On the other hand, since the weight emphasizes the pair of classes that are close together and deemphasizes those that are far away from each other, the proposed method will be more advantageous for the classification task. The experimental results on real face database confirm the better recognition performance of the proposed method.