Synthetic Sentiment Energy: Situation Assessment in Crowds

Group Emotion Recognition has been recognized as a difficult task because of the difficulty in defining identification elements, the overlap of people, and the differences of individuals. This paper first discusses the two elements required for group emotion recognition: The facial expression on each person’s face and the position of each person’s face in the crowd. Then, it design Binary-Channel CNN and Yolo V3 respectively as a strategy to obtain these two kinds of information. This paper propose a Sentiment Energy Function to measure group sentiments, which consists of three components: Expression energy, Expression Change Rate Energy and Displacement energy. These three components were fused aiming to achieve the group emotion energy. Through experiment methods, it is clear that the improved Binary-Channel CNN in this paper perform good in facial expression recognition. The group emotion recognition system designed has comparatively higher robustness and adaptability. The results obtained are consistent with the results of human analysis.