Sunnfering understanding system based on facial expression and human action

An emotion understanding system is proposed for SUFFERING factors to interpret the negative emotions, since human's emotion cannot be completely represented only by facial expressions or actions, respectively. The proposed system mainly composed of facial expression recognition and action detection, in which three methods are contributed in this work. Firstly, 19 Suffering Units extended from Ekman's Facial Action Units System are proposed for negative emotion detection and happiness promotion. Next, the proposed Hierarchy-Coherence K Nearest-Neighbor calculates the coherence for facial expression classification, the intra-class coherence is considered as hierarchical nearest neighbors to benefit the KNN classification. Finally, a three-layer Average Moving Action Status Window is presented to improve the computational efficiency for action detection. Furthermore, a Voting Look-up Table is used for comparison between suffering factors and suffering units. The experimental results have demonstrated the effectiveness of the proposed system, the accuracy rates achieve 96.98% and 89.94% for facial expression and action detection, respectively.