The happiness, fear, sadness, anxiety, and etc. feelings appear in the face for a short time after controlling and suppressing. This unmanageable expression is micro-expression. It can show trickery and lie. Therefore, its recognition has a wide of applications in tribunal and therapy. However, it is necessary to identify the rapid and weak movements of the facial muscles. To this end, in this paper, we magnify motions and we have considered six planes along the time dimension to identify changes well. We also propose the multi-color uniform local binary pattern from six intersection planes and the histogram of gradient direction from these planes, which produce good results in micro-expression recognition. These methods can be used in a real environment. Because they work better in illumination and light changes. The result of the experiments shows 74.12%, 86.19%, and 66.23% accuracy using the proposed method on CASME I, CASME II, and SMIC-NIR databases, respectively. Thus, our method has promising performance in comparison with the methods of previous researches.

Keywords: Micro-expression recognition, histogram of gradient direction, multi-color uniform local binary pattern

Full-Text [PDF 3077 kb]   (743 Downloads)