Gender classification with jointing multiple models for occlusion images

A facilitated and effective gender recognition approach is desirable for various applications such as for intelligent surveillance systems, human-computer interactions, and consumer behavior analysis. Since the human face conveys clear sexual dimorphism, the use of facial features seems an intuitive way to recognize gender. This paper proposes an efficient gender classification method using multiple classifiers to overcome the occlusion problem. The experiment is tested via 5-fold cross validation on the FERET and AR databases to evaluate the performance. The results show the proposed approach achieves higher accuracy than previous methods.