Facial expression recognition in video sequence based on LBP feature and GRU
Luo Lin, Shengwei Qin, ZiLong Wu, BingQuan Xu · 2021
Human facial expression can convey rich information, let the computer understand the human intention, and make correct responses through facial expression recognition, which is a hot spot in artificial intelligence research at present. Compared with a single frame image, a video sequence image contains the information of expression changing along the time axis, which can provide more help for expression recognition. Therefore, this paper proposes an expression recognition method based on video sequence images. Firstly, the facial landmark points are extracted from the sequence images. Secondly, four key regions that contribute greatly to expression recognition are calculated through the landmark points. Then the LBP features of these four regions are calculated and fused to form the expression features of a single frame image. Finally, these expression features are sequentially sent to the GRU network for training to obtain the face expression classification model. The experimental results show that the proposed algorithm has high recognition accuracy.