一种基于运动姿态的三维人体运动检索方法

Motion Retrieval Based on Motion Posture

随着运动捕获技术的不断展,捕获出大量的三维人体运动数据并建立了大规模运动数据库。提出了一种基于运动姿态的三维人体运动检索方法(MPDS),首先进行姿态特征提取,其次通过特征编码对人体运动的局部姿态进行组合,以期达到更好地辨别力,并通过决策树学习方法得到局部运动姿态相似度对整体运动相似性的影响,以决策树指导检索过程。实验结果表明,文中算法的有效性。

With the development of motion capture technique, a lot of motion data and libraries become available. In this paper, we presented a novel method for motion retrieval based on motion posture（MPDS）. First, motion pos- ture features were extracted with the help of human motion characteristic. Then, the feature coding function was pres- ented and the decision tree was automatically constructed to determine movement similarity by local movement pos- ture. Experimental evaluation of the method show that the proposed method is effective for motion retrieval.