您的位置:首页  > 论文页面

基于稀疏约束的人体全身运动合成方法

发表时间:2017-08-31  浏览量:1792  下载量:528
全部作者: 刘华俊,杨青叶,何炎祥,张沪寅
作者单位: 武汉大学计算机学院
摘 要: 利用稀疏控制信号对高维人体运动动画进行精确控制的一个难点在于低维控制信号本身难以准确且唯一合成高维人体运动。针对此瓶颈问题,提出一种新颖而强大的局部模型降维方法,将提前捕获的高维人体运动数据进行降维处理,并将其用于自然人体全身运动的准确合成。该方法从预先捕获的运动数据库中构建一组在线局部动态的回归模型,以支持高维人体全身运动的重构。与以前的统计约束模型相比,本研究提出的模型可以合成更加准确的结果,实验也证明了所提方法的有效性。
关 键 词: 计算机仿真;数据驱动动画;稀疏信号;人体运动合成
Title: A sparse constrained method for full-body human motion synthesis
Author: LIU Huajun, YANG Qingye, HE Yanxiang, ZHANG Huyin
Organization: School of Computer, Wuhan University
Abstract: One difficulty for constructing high-dimensional human motion animation precisely by using sparse control information is that low-dimensional control signals cannot be used to synthesize high-dimensional human motions. We introduce a novel and powerful local dimensionality reduction approach for constructing natural and accurate full-body human pose sequences by dimensionality reduction method based on pre-recorded high-dimensional human motions. Our approach is to build a group of online regression models from a pre-recorded motion database to support the reconstructing of high-dimensional human action. Compared with the former statistical models, our approach can construct more accurate results, and the effectiveness of the model has been verified.
Key words: computer simulation; data-driven animation; sparse signal; human motion synthesis
发表期数: 2017年8月第16期
引用格式: 刘华俊,杨青叶,何炎祥,等. 基于稀疏约束的人体全身运动合成方法[J]. 中国科技论文在线精品论文,2017,10(16):1868-1876.
 
1 评论数 0
暂无评论
友情链接