Abstract: Based on 3D features for time and space relationship of each human joint，a kind of temporal-spatial features was proposed in this paper, which can process local temporal-spatial features of each joints. 3D features can represent a part of the whole motion independently and avoid contacting with original motion data and eliminate "curse of dimensionality". Experimental results show that the approaches are effective for motion data analysis, which can be used in intelligent analysis for motion semantics.

Key words: motion capture, 3D feature, temporal-spatial feature

摘要： 基于各关节点三维空间和时间关系的三维特征，提出了一种描述各关节点之间三维空间关系的时空特征，可以分别处理每一个关节点的局部时空特征。同时，三维时空特征避开与原始数据的直接接触，从而很大程度避免了维数灾难。实验结果显示，三维特征提高运动数据检索的效率和精度，可应用到运动语义智能分析等领域。

关键词: 运动捕获, 三维特征, 时空特征