Design And Inplementation Of Face Recognition Algorithm Based On FPGA

With the explosion of information today, the information security has becomed an important issue. As one of the most important technology in the recognition by the biometrics, compared to other recognition method, like finger recognition, Gene recog-nition,speech recognition and iris recognition, face recognition is convenient, traceable, and also it has high recognition rate. Face recognition is widely used in national security, security and entrance and exit detection and so on. The technology of face recognition is very mature on PC. With the birth of the electronic consumption, it sometimes doesn’ t meet the demand of the embedded SOC. Also DSP can’t have the capability of pro-cessing the video data and image data due to the limit of the logic source. With the fast development of the performance and integration of FPGA in the recent 20 years, it makes the application of video and image with FPGA possible. FPGA has many ad-vantages, such as many logic resources, high speed, high flexibility. What’s more, the FPGA firms offer the FPGA chips embedded ARM, users can easily design circuits with the cooperation of software and hardware. Processing the video and image with FPGA becomes a trend now.The process of face recognition is mainly composed of face detection and position, image preprocess, feature extraction and face matching. In the part of face detection and position, this paper uses the face detection based on the skin color, and then with the morphological filtering and other method, we simply find the position of the face. In the part of image preprocess, the paper realize the Gaussian filter, median filter, histogram equalization, and canny edge detection algorithm with FPGA. In the part feature ex-traction, first, the paper realize principle component analysis, independent component analysis, nonnegative matrix factorization, and discrete cosine transform algorithms, and then apply these four algorithms to the ORL database, second, the paper analysis the affection of the degrees of the feature vector and the number of the training set to the recognition rate, and the results shows that the discrete cosine transform has the highest recognition rate on the ORL database compared to the other algorithms, which can be up to 97.5%. Last, the paper realize the discrete cosine algorithm with FPGA, and pass the feature vector to the ARM by the AXI protocol, then match it with the face database.This paper realize a simple face recognition system, and applied it to the people in my laborratory. The rate of recognition can reach to 91.6%.