中文摘要：

如何通过指定约束条件的方式交互式地合成风格化人体运动是计算机动画研究领域的热点和难点,传统的数据驱动办法通常没有全面考虑运动的静态和动态特性.针对这一问题,文中提出人体运动的独立时空特征空间模型,利用一个可变形运动模型和独立特征子空间分析算法提取运动在时空两个域上的特征,并将其封装起来,通过低维子空间进行描述.运动风格的编辑可利用低维运动混合和空时约束优化等方法来实现.该方法充分考虑了运动在时域上的特征,并为用户提供了可编辑接口.文中以各种风格的行走运动为例讨论了该方法的有效性,结果证明效果良好,可用于交互式运动风格编辑.

英文摘要：

It is still a challenge to generate interactively stylistic human motions satisfying userdefined constraints in computer graphics community. Previous works failed to generate stylistic human motions with both geometric and timing features. As a solution to this problem, a motion model called Independent Spatio-temporal Feature Space is presented in this paper. This model encapsulates spatio-temporal features for motion style which can be extracted from a deformable motion model with an unsupervised machine learning techniques-independent feature subspaee analysis. The spatio-temporal features are described as several low-dimensional subspaces. Motion style editing is subsequently achieved by blending low-dimensional parameters or solving space-time constrained optimization. Both geometric and timing features are considered in this work and a directly interactive interface is provided. The power and flexibility of this method are demonstrated by taking locomotion as examples. The experimental results show that this method is fast and effeetive for interactive motion style editing.