基于人体姿势编码的运动数据检索

Motion Data Retrieval Based on Human Pose Coding

针对大规模运动捕获数据的检索问题,提出一种基于人体姿势编码的相似运动检索方法.首先将形成人体姿势的9段主要骨骼的运动空间划分成多个子空间,以各骨骼所处的子空间位置作为人体姿势的特征表示;然后根据特征表示对运动数据进行逐帧编码,将编码相同的连续帧划分成等价段;最后建立以等价段为基础的倒排索引来检索候选运动,并通过编码匹配算法确定最终检索结果.实验结果表明,该方法具有较好的检索性能,能够实现相似运动的高效检索.

This paper presents a novel approach for retrieving similar motions from large-scale motion capture database with human pose code. Our approach employs nine major bones to represent human pose. First, each bone＇s active space is partitioned into several subspaces, whose positions are used as feature representations of human poses. Then, a motion data is encoded frame by frame with the feature representation, and consecutive frames which share a motion code are grouped. Thereafter, grouped frames are assigned an inverted index for retrieving representative motions. Similar motions are extracted from these representative motions by code matching algorithm. Experimental results show that our approach can effectively retrieve the similar motions from motion capture database.