The Interaction Ontology Model: Supporting the virtual director orchestrating real-time group interaction

In a system that enables real-time communication between groups of people via audio/video streams, a component called orchestration intelligently selects appropriate camera views for each participant individually, enabling larger setups and enhancing the social interaction itself. The Interaction Ontology (iO) receives low-level cue input from the audiovisual analysis component and informs the camera view switching component about the social interaction on a higher semantic level. The iO is a software component consisting of both a static ontology model and dynamic event processing logic. In this paper, we elaborate on the design rationale of the model and the intended dynamic behaviour of low-level cue processing. Finally we discuss performance and scalability issues, as well as alternative approaches to low-level event processing in such environments.