System and Method for Visual Analysis of Emotional Coherence in Videos

摘要

A computer implemented method and system processing a video signal. The method comprises comprising the steps of: detecting a human face displayed in the video signal and extracting physiological, biological, or behavior state information from the displayed face at a first level of granularity of the video signal; processing any two or more of: (i) a script derived from or associated with the video signal to extract language tone information from said script at a first level of granularity of the script; (ii) an audio signal derived from or associated with the video signal to derive behavior state information from said audio signal at a first level of granularity of the audio signal; (iii) a video image derived from the video signal to detect one or more human gestures of the person whose face is displayed in the video signal; and merging said physiological, biological, or behavior state information extracted from the displayed face in the video signal with any two or more of: (i) the language tone information extracted from the script; (ii) the behavior state information derived from the audio signal; and (iii) and the one or more human gestures derived from the video image, wherein the merging step is based on behavior state categories and/or levels of granularity.