Facial expression recognition via Gabor wavelet and structured sparse representation

Automatically facial expression recognition (FER) has become more and more important today, making machine understand human's emotion by expression having various potential applications, especially in the field of human machine interaction (HCI). But FER still remains a challenge problem in computer vision as the subtleness of facial expression is difficult to capture and the robustness of the recognition in various situation is also hard to guarantee. In this paper, a Gabor wavelet and structured sparse representation based classification (SSRC) are proposed aiming to solve the FER problem. The Gabor wavelet filter is used to extract the subtle facial expression, and the structured sparse representation based classification (SSRC) is used for classifying the test images robustly. Unlike sparse representation based classification (SRC), the SSRC explicitly takes structure of the dictionary into account for a better classification. Experimental results show the better performance of our proposed method compared with other traditional methods, especially more robust in case of facing corruption or occlusion.