With the development of convolutional neural network, the accuracy of face recognition has been significantly improved. Currently, face recognition systems are mostly based on CPU or GPU, which are difficult to perform well on embedded devices. However, face recognition algorithms based on deep learning in embedded devices are often limited by computing power. This live demonstration presents a face recognition system based on ARM+FPGA heterogeneous computing system. Face detection and face alignment are completed on ARM, and face recognition are completed on FPGA. Face recognition algorithm based on convolutional neural network is designed according to the hardware characteristics of FPGA. The whole system can reach the speed of 93 ms, in which the speed of face detection is 45 ms, the speed of face alignment is $38$ ms and the speed of face recognition is $10$ ms. At the same time, face recognition on the LWF data set can reach 99.05\% accuracy.