Motion retrieval based on multi-view information and graph model

Matching and retrieval of motion sequences has become an important research area in recent years, due to the increasing availability and popularity of motion capture data. In this paper, we propose a novel content-based method for retrieving of this human motion captured data. The proposed method includes two key steps, one is motion descriptor construction, and the other is retrieval based on query. Firstly, for representing motion by images, multi-view information is captured, and then switching liner dynamic system (SLDS) is built based on image fusion and light stream technology. Secondly, through inferring and coding of SLDS, a graph-based motion descriptor can be obtained. Lastly, based on proposed motion descriptor, the motion retrieval would be operated through direct matching between motion descriptors. Experiment results show our method is effectiveness.