The way that people move together, for example when dancing or sporting, carries semantic and emotional information. In this work package in Virtual Worlds for Wellbeing, we will visually analyze video sequences of groups of people in motion, in order to extract such information.
The objectives of this work package is to perform research on aspects of the extraction of distinguishing features to describe and categorize motion styles, arousal, happiness, etc. The methods to be developed must be generic, in the sense that they are applicable under various conditions such as lighting, angle of view, number of people, etc.
Video analysis of people has been done in many areas, for example security, sports, and traffic. More specifically, the analysis of motion of people has been analyzed for training purposes in sports and dance, and for making synthetic animations more realistic. These research results apply usually to individual people, not the groups of people. In this work package we perform research on the video analysis of motion of groups of people. We will generalize earlier work, including our work at UU, about motion and gesture recognition, to the emotional interpretation of movements. This will be done primarily on the basis of video, but we will use our motion capture lab to obtain referential data, as we have done earlier in other domains.