Facial Expression Recognition Based on Video

In recent years, video timing characteristics of facial expression recognition research has become a hot topic. In this article, through analyzing changes in the feature data in realtime detection of facial expression, and combining temporal and spatial features to establish models, we put forward a more rapid and more efficient method to recognize facial expressions. Firstly, the feature extraction method based on Bezier curve is adopted to extract the 2D feature points of each frame in the video stream. Then, we get the changes of temporal characteristic curve through connecting the feature points of each frame image along time slice. Finally we use nonlinear function to fit the changes of the temporal characteristics curve and we establish models for each expression to classify and recognize. The results show that the method which combines the temporal and spatial changes in facial expression recognition, has the advantages of high speed and recognition rate.