A Research Of Emotion Recognition Based On Dynamics Facial Expression In Eastern

Emotion is an important part of human psychological reaction. Emotions can accurately convey the human thoughts. So it is essential for human facial expression communication in People’s Daily communication. For a long time, the issue of the universal hypothesis of facial expressions has become a hot topic. Some researchers believe that people in different cultures have the common emotion cognition, which has the same six basic expressions(happy, surprise, fear, disgust, anger, sadness); Other researchers is based on the research of cross-cultural direction, they refuse the emotion universal hypothesis and confirm that emotion has the characteristic of cultural diversity.Now many studies of emotion are based on the static image. However, the dynamic facial expressions transmit a very wealth of signal. We use dynamic facial expressions code system to collect experimental data of 61 Chinese subjects. Then we built subjects’ perception model of dynamic facial expression. So we can study the dynamic cognitive processes of six emotion for Easter. The results show that the facial expressions cognition for Easter has following features.(1) "surprised", "fear" and "disgust" in Easter have confusion significantly.(2) In Eastern, early face signals extinguwish emotion for four categories or five categories. Late face signals support discrimination of all six emotion categories. So the transmition of the dynamic facial expressions code system has culture specific and the expression of emotion is not universality.In this paper, we further propose a framework for facial expression recognition based on the specific facial expression(i.e., the corresponding Action Units of six basic expressions) of Eastern. To recognize certain facial expression, we extract the features within the limited regions of the specific action units corresponding to the target facial expression. Experimental results show that our method improves the accuracy of facial expression recognition by extracting features from the corresponding local regions(action units). It is also concluded that the cognitive mechanisms of facial expression will inspire the algorithm design for facial expression recognition or other intelligent interaction systems.