Micro-expression recognition from local facial regions

MiE is a facial involuntary reaction that reflects the real emotion and thoughts of human being. It very difficult for normal to detect Micro-Expression (MiE), since it fast local face with low intensity. As consequence, challenging task researchers build an automatic system recognition. Previous works recognition have attempted use whole face, yet appears in small region which makes extraction relevant features hard task. In this paper, we propose novel deep learning approach leverages locality aspect MiEs by spatio-temporal from regions using composite architecture Convolutional Neural Network (CNN) Long Short Term Memory (LSTM). The proposed solution succeeds extract Experimental results on benchmark datasets demonstrate highest accuracy our respect state-of-the-art methods.