摘要
逆运动学(Inverse Kinematics)是虚拟角色运动控制的一种基本方法,它根据用户指定肢体末端的位置计算出虚拟角色各个关节的旋转.传统算法求解时没有考虑人体姿态的运动规律,因此其结果不能完全令人满意.文中提出了一种利用捕获的运动数据辅助求解逆运动学问题的新方法.通过自组织映射(Self-Organizing Map,SOM)对姿态数据学习和聚类,获得一组刻画人体姿态空间的支撑姿态,然后通过对问题所在局部空间的支撑姿态加权优化来求解逆运动学问题.该方法克服了传统方法结果不自然、计算效率较低的缺点.实验结果也表明了该文方法的有效性.
Inverse kinematics(IK) is one of the most frequently used methods in motion control. It is used to calculate the joint rotations of articulated figure given the end effectors' positions. Traditional methods can't produce the natural results because of the lack of consideration in the law of the human motion. This paper presents a novel method to solve IK problem with the assistance of motion capture data. The SOM(Self-Organizing Map) is adopted to obtain a set of reference poses from motion capture data and search for the solution in the local space which is spanned by weighting the local reference poses. The experiments show the robustness and effectiveness of the proposed method.
出处
《计算机学报》
EI
CSCD
北大核心
2007年第11期1982-1988,共7页
Chinese Journal of Computers
基金
国家自然科学基金(60533070
60603082)
国家"八六三"高技术研究发展计划项目基金(2006AA01Z336)
北京市自然科学基金(4062032)资助.~~
关键词
角色动画
逆运动学
单位四元数
自组织映射
优化计算
character animation
inverse kinematics
unit quaternion
self-organizing map
optimizations