In this paper, we present a video face recognition system and study a face recognition method which uses feature fusion in video. The characteristics of LBPH, SIFT and VIPLFaceNet are firstly discussed. LBPH is a feature which is based on the local gray difference of image. SIFT is a kind of feature which is invariable to scale. VIPLFaceNet is a kind of algorithm and feature which uses convolution neural network. The fusion of these three features can eliminate the influences of illumination, rotation and scale change on recognition. For the experiment, the ORL face data set and a real video was used to test, and the results show that, compared with single feature recognition, recognition rate has improved significantly with feature fusion.