Maximum similarity degree for 2D fuzzy face recognition

Abstract

In this paper, a maximum similarity criterion is proposed which is adapted to a new fuzzy face recognition method (namely, 2DFMS). The similarity degree between faces is defined by a nonlinear function. Based on this similarity, an improvement fuzzy membership function is obtained by applying k-nearest neighbor. Then, 2DFMS extracts the features from face images directly so that it will not suffer from the SSS problem. Finally, in the projected space, the test image is identified according to a specific classifier, which is based on a maximum similarity criterion. The whole algorithm is implemented on ORL and Yale face database to demonstrate the effectiveness and robustness.