Facial Expression Recognition Based on Local Vector Model

Texture feature extraction is an important step inthe facial expression recognition system. The traditional LBPmethod ignored the statistical characteristics of the texturechange direction in the process of feature extraction, and wecan extract more detailed texture information by the LDPmethod based on LBP, but the computational complexityis greatly increased. In order to extract more detailedtexture information with the computational complexity is notincreased, we proposed a method named Local Vector Model(LVM). In this method, modulus value and direction of thelocal texture changes are extracted as the features of classification. Furthermore, in order to improve the robustness thatthe algorithm to the subtle deformation of expression image,the Image Euclidean Distance is introduced and embeddedin LVM. Finally, the even decreasing function is used toget the neighbor classification distance. Experiments onJAFFE facial expression databases with different resolutiondemonstrated that the method proposed in this paper isbetter than other modern methods.