Activity recognition through multi-scale dynamic Bayesian network

Activity recognition is one of the most challenging problems in the video-based surveillance and computer-vision. In this paper we propose a novel approach to recognize human activity in which we decompose an activity into multiple stochastic processes, each corresponding to one scale of motion details. We present a hierarchical durational-state dynamic Bayesian network(HDS-DBN) to model two stochastic processes which are related to two appropriate scales in intelligent surveillance. In this approach the features we extracted are divided into two classes: global features and local features, which are at two different spatial scales. The HDS-DBN model structure combines global features with local ones harmoniously. The effectiveness of our approach is demonstrated by the experiments.