Video Facial Expression Recognition Based on Multi-angle Feature Fusion

Abstract

In view of the problems of high computational complexity and low recognition rate of video facial expression recognition, this paper proposes a new method of video facial texture representation and facial expression recognition, which is based on multi-angle feature fusion algorithm of local Gabor adaptive ternary orientation pattern (LGATOP). Firstly, in order to obtain an enhanced Gabor amplitude spectrum, Gabor wavelet transform with different scales and directions is applied to each face image normalized in the video frame set. Then the LGATOP features of the video texture are extracted by using the local three-value model on three orthogonal planes. Finally, the SVM is used to complete the emotion classification of the texture and motion information after fusion. Through the cross experiment on CK+ and MMI databases, it has been proved that the proposed method has better robustness to mination change, expression changes and so on. Contrast experiments with other related methods have also verified the superiority of the method.