Abstract
In order to solve the problems of low efficiency in facial expression recognition and difficulty in engineering implementation, this paper presents an infrastructure for a facial expression recognition video analysis system and method, centered on emotion computing services, relying on Hikvision's facial action unit analysis server, so that the capabilities of expression analysis service are provided to upper-level applications. It focuses on the construction methods of the emotion rule library and the facial action unit weight vector library, proposes an expression recognition and emotion computing core algorithm based on the highest confidence priority of multiple matching rules, uses vector operations to achieve fast matching of multiple rules, and introduces a simplified emotion state quantification method, which is used to measure the intensity extents of emotions. And then it designs the detection method of emotion fluctuations and the corresponding methods of emotion fluctuation confidence and intensity quantification.