Face Recognition Based On Deep Learning

With the development of computer information technology,biometric identification technology has been paid more and more attention.Face recognition has become one of the most important techniques for biometric recognition because of its non-contact and security.More and more experts and scholars have begun to study the technology of face recognition,and have made a lot of achievements.In this paper,based on the research of deep learning technology,a new method based on convolution neural network F-CNN has been proposed.The F-CNN model combines SN neural network and B-CNN neural network,which improves he high-level layer size of the neural network,and enhances the ability of expressing the depth CNN network model.Experiments were carried out in the international face database Yale Face,the experimental results show that the fusion method of convolutional neural network capability is presented in this paper is higher than that of SN network and B-CNN network,the recognition rate reaches 98.14%.With the commercial application of face recognition system,static picture,video and 3D mold as the representative of the deception can easily deceive the general system of face recognition,which poses a great threat to the application of face recognition system.Therefore,this paper proposes a detection method based on in vivo instruction control.The basic principle is that the faces in the static image around the face area almost have no change in the process of de:flection,but the living face deflection in the process,the left and right face area ratio will change a lot.Through the experiment of the 50 live users and their face print images,the result proves that the method has good detection effect on anti static pictures,False Accept Rate FAR reaches 4%,False Reject Rate FRR reaches 0,Comprehensive Accuracy Rate CAR reaches 98%,the test results are better than the traditional Fourier spectrum analysis method and optical flow method.