Facial expressions recognition based on joint shape-texture cues in video

Abstract

This article proposed a facial expressions recognition system, which has a high recognition rate and can work in real-time, in frontal pose video. Our main contribution is that we present a new so-called ASL descriptor which is combined from three different basic features extraction method: AAM, SURF and LBP. Extensive experimental results demonstrate that our method achieve the performance higher than some state-of-art results on the Cohn-Kanade Database (CK).