基于KDT的人体运动数据行为分割
首发时间:2014-03-11
摘要:针对人体运动数据的行为分割问题,描述了一种基于核动态纹理KDT(Kernel Dynamic Texture)的行为分割方法。这种思想借鉴了动态纹理在视频图像处理领域上的应用。该方法主要包括以下三个步骤:首先针对人体运动数据预处理,选取参考序列片段和子序列片段;接着利用核PCA(Principal Component Analysis)方法,对这些片段构建KDT;最后基于马丁距离度量不同人体运动数据KDT之间的相似性,通过相似性距离的跳变特性检测人体运动数据的行为分割点。通过实验证明了所描述的行为分割方法在准确率和召回率指标上表现出了良好的性能。
关键词: 计算机应用 行为分割 核动态纹理 运动捕捉 角色动画 人体运动分析
For information in English, please click here
Behavioral Segmentation for Human Motion Data Based on Kernel Dynamic Textures
Abstract:For the task of behavioral segmentation, This paper presents a motion segmentation method based on the Kernel Dynamic Texture (KDT), which is inspired by video and image processing with dynamic textures. The proposed approach mainly includes the following three steps: First, the human motion data is preprocessed. The reference sequence and subsequence clips are selected. Second, these clips are constructed KDT with kernel Principal Component Analysis (PCA) method. Finally, Martin distance is exploited to measure the similarity of distinct human motion data KDT, and we detect the behavioral cuts through analyzing the jumping features of similarity distance. Experiments show that the presented method performs good in term of recall and precision rate.
Keywords: computer application behavioral segmentation kernel dynamic texture motion capture character animation human motion analysis
基金:
论文图表:
引用
No.4588278961250139****
同行评议
共计0人参与
勘误表
基于KDT的人体运动数据行为分割
评论
全部评论0/1000