A Hybrid Fuzzy Semi-supervised Learning Algorithm for Face Recognition

In this paper, we develop a hybrid fuzzy semi supervised learning algorithm (HFSA) for face recognition, which is based on the segregation of distinctive regions that include outlier instances and its counterparts. First, it achieves the distribution information of each sample that represented with fuzzy membership degree, and then the membership grade is incorporated into the redefinition of scatter matrices, as a result, the initial fuzzy classification of whole regular feature space is obtained. Second, a new semi-supervised fuzzy clustering algorithm is presented on the basis of the precise number of clusters and initial pattern centers that have been previously obtained in the pattern discovery stage, and then applied in order to perform the outlier instances classification, yielding the final pattern recognition. Experimental results conducted on the ORL and XM2VTS face databases demonstrate the effectiveness of the proposed method.