A Fast Face Recognition System Based on Deep Learning

With the advent of the era of big data, deep learning theory has been rapidly developed and applied, especially in the field of image recognition. Compared with the classic recognition algorithm (such as LBP [1] and PCA [2] algorithm), deep learning algorithm has the characteristics of high recognition rate and strong robustness. Based on the principle of convolution neural network [3] (CNN), a realtime face recognition method on FPGA was proposed, which improves the speed and accuracy of face recognition. The method is divided into two parts. First, the PC terminal is used to complete the training of the network and get the network parameters. Secondly, the face recognition system is built on the FPGA. The advantage of FPGA parallel processing is to speed up the computation speed of the network so as to achieve the purpose of real-time processing of face recognition. The test results showed that the recognition speed of the system has reached 400FPS, far exceeding the existing results. The recognition rate is 99.25%, higher than the human eye. Moreover, it has good robustness for complex light environment.