Research On Gender Recognition Based On Face Images

With the continuous development of Computer Intelligence technology, people pay more attention to face recognition.Gender recognition technology which is an important part of the face recognition technology, also needs more and more high recognition rate and robustness.Gender recognition technology with facial image is to find the face from the input image and distinguish its gender. The face feature extracton and gender classification determine the result of the gender recognition. Researchers raised many excellent feature extraction algorithms and classification algorithms to the framework of gender recognition. The algorithms have achieved some results, but there are still many issues remain to be solved, such as illumination variety and the small occlusion of face.The thesis was started to study the above issue, and the main research parts are as follows.1, The thesis studied the various methods of gender identification, and chosed convolution neural networks as the main research method.Meanwhile, it resarched the network structure of convolution neural network, forward propagation algorithm and back propagation algorithm.In order to decrease the number of the training of the network, it also studied the back-propagation algorithm based on Fisher criterion.2, The face detection method based on AdaBoost is the most popular face detection method,this paper chosed it as the face detection method. This method often gets some error results, so the thesis chosed Active Appearance Model to verify the detection results.Many methods were carried out in pre-treatment to standardize the images, such as, gray image, histogram equalization, normalization and face correction.3, In order to obtain a method which has high recognition rate and robust, the thesis researched the principle of Gabor features.Then the author came up with a gender recognition algorithms which combines Gabor features and convolution neural networkbased on Fisher criterion and analysised this method.Finally, the paper designed a gender recognition system based on this method.4, Many experiments had been designed to verify the effectiveness and robustness of this method.Such as different face databases, different gender recognition algorithm, different light and the small obscur of face.These experiment sproved that the algorithm has high recognition rate and robustness. In order to explain the application value of this method, this paper designed a gender recognition algorithm platform.