Research And Implementation Of Gender Recognition Based On Face Image

This thesis mainly studies the method of gender recognition based on face images. The external face feature and hair feature are combined to decide gender, and in classification stage, classifier fusion is realized with a parallel structure of weighted voting algorithm to enhance the classification accuracy.Firstly, an improved injection filling algorithm is proposed to detect the area of hair. Then a model is established to extract the features of length, thickness and area of hair zone. Secondly, facial geometric feature called active shape model and local binary pattern are extracted. Next, this thesis presents a parallel structure of weighted voting algorithm to realize the classifier fusion. At last, we describe the environment configuration and the overall design of the gender recognition system, and then give the detailed design and implement for each module. Experiments show that the method to combine hair feature and facial feature and use multiple classifier fusion performs well.