Context analysis in intelligent meeting scene

Context analysis is a crucial issue in dynamic meeting scenarios for online information service and offline semantic information retrieval. Four high-level semantic contexts, that is, "monologue", "presentation", "discussion" and "break", are considered and analyzed in dynamic meeting scenarios in this paper. To characterize four semantic contexts by event fusion and inference, four kinds of group events are considered, such as participant\´s motion events, speaking events, appearance of person events, and using the projector screen event. Hierarchical dynamic Bayesian networks, which model group events and situation context, are constructed to bridge the gap between physical audio features and semantic concepts. Rao-Blackwellized particle filter is applied for online inference in our hierarchical dynamic Bayesian network. The experimental results demonstrate the effectiveness of the proposed approaches.