基于关键特征的运动搜索方法
首发时间:2013-04-02
摘要:为了实现快速的运动搜索,本文提出一种基于关键特征的运动搜索方法。该方法应用时空Isomap对运动捕捉数据进行降维,从而保留运动的关键特征;然后对运动的低纬嵌入进行简单的迭代对关键运动特征所在的关键帧进行定位;并动态的建立一个二叉关键特征树。实验证明经过少量的预处理时间,可以获得实时、有效的运动搜索结果。
关键词: 多媒体检索 运动检索 关键运动特征 二叉关键特征树
For information in English, please click here
Motion Retrieval Based on Key Features
Abstract:In this paper a fast motion retrieval method based on key motion features is proposed. The method applies the Spatio-Temporal Isomap (ST-Isomap) on effective aspects of motion data to yield a reduced low-dimension space, where the main characteristics of motion are kept. The lower complexity of the embedded motion curve allows a simple iterative method which locates the key-poses of the original motion. We present a dynamic method of constructing a BFS Tree with significant points to search for a given motion example. With proper dynamic algorithm and key-pose selection in terms of spatio-temporal variations, the method can consume a small amount of pre-processing time, and achieve precise and effective retrieval.
Keywords: multimedia retrieval motion retrieval key motion feature BKS-tree
论文图表:
引用
No.****
同行评议
共计0人参与
勘误表
基于关键特征的运动搜索方法
评论
全部评论0/1000