Motion retrieval based on temporal-spatial features by decision tree

Abstract

In order to retrieve similar motion data from Mocap database, each human joint's motion clip is regarded as a bag, while each of its segments is regarded as an instance. 3D temporal-spatial features are extracted and data driven decision trees are automatically constructed to reflect the influence of each point during the comparison of motion similarity. At last we use the method of multiple instance retrieval to complete motion retrieval. Experiment results show that our approaches are effective for motion data retrieval. © Springer-Verlag Berlin Heidelberg 2007.