Gender Recognition Based On Facial Image

Sex is the most basic information to classify each other for people, usually we can distinguish a person's gender intuitively and easily,however, it's difficult to a computer. In recent years, with the development of computer science and technology, the function of computer has become more and more powerful and intelligent. The imitation of various human biological characteristics has become a hot research topic in the computer industry. As an important research topic of intelligent human-machine interaction, facial gender recognition technology has been widespread attention and rapid development in recent years,and it has important theoretical significance and application prospects in authentication, video retrieval and robot vision. Usually, Facial gender recognition system is divided into face detection and pretreatment, face feature extraction and gender classification three parts. In this paper the three aspects were studied. The main work of this paper are as follows:(1)Summarize the purpose and the significance of facial gender recognition. Especially for gender feature extraction and gender classification method of two parts, the status of research are given by this paper. By reading papers, summarize the latest development of both gender feature extraction and gender classification.(2)In this paper, face detection technology was reviewed into three classes, after the face detection and eyes localization of input face images, the mapping results of eyes localization were used to correcting the oblique images. For the face detection of face image,this paper introduced the theory of Adaboost method and researched positions of two eyes on the basis of the Adaboost method, then the accurately positions of two eyes were located fastly and accurately through block integral projection. After determining the eyes location,the calculation method of the rotation angle was given with the geometric center of image as rotation reference point,and achieved rotation correction,it has certain significance for perfecting the image correction theory.(3)Introduced the basic theory of active appearance model, including the establishment of shape, texture and appearance model, and the calculation of image fitting. In the process of fitting,the fitting principle and steps were given respectively, and the changing process of image in iterative process was given by experiment. By fitting the tilt face image, the practicability of rotation correction of the tilt face image in chapter second was further confirmed, the shortcomings of the AAM algorithm and the limitation of use were analysis finally.(4)A precise patch histogram was adopted to use for gender classification. Based on the localization of facial feature points, this paper adopted a method of extracting feature patch for facial feature points, and constructed the patch histogram based on the LBP algorithm, the histogram distinction between male image and female image were used to gender classification through Support vector machine(SVM).The best case of patch in gender classification and the worst case of patch were analysed in result, and it showed that the method was robust in the accuracy of gender classification.