Representation of Motion Spaces Using Spline Functions and Fourier Series
Representation of Motion Spaces Using Spline Functions and Fourier Series
复制标题
使用样条函数和傅立叶级数表示运动空间
DOI：
--
复制
发表时间：
2012
期刊：
影响因子：
--
通讯作者：
G. Brunnett
中科院分区：
文献类型：
--
作者：
Thomas Kronfeld;Jens Fankhänel;G. Brunnett
由于运动数据的高维性，看起来自然的人体运动难以创建和操纵。在过去的几年里，大量的运动捕捉数据被用来增加角色动画的真实感。为了简化运动的生成，我们提出了一种数学方法来创建运动数据的变化。给定特定活动的运动数据的一些样本，我们的框架生成高维连续运动空间。因此，我们的运动合成框架是能够通过不同的边界条件合成运动。此外，我们调查的不同性质的样条函数和傅立叶级数和它们的适用性的描述复杂的人体运动。我们已经推导出一个优化启发式，这是用来自动生成初始运动空间。我们通过与地面实况运动数据和替代方法的比较来评估我们的系统。
Natural looking human motion are difficult to create and to manipulate because of the high dimensionality of motion data. In the last years, large collections of motion capture data are used to increase the realism in character animation. In order to simplify the generation of motion, we present a mathematical method to create variations in motion data. Given a few samples of motion data of a particular activity, our framework generates a high dimensional continuous motion space. Therewith our motion synthesis framework is able to synthesize motion by varying boundary conditions. Furthermore, we investigate the different properties of spline functions and Fourier series and their suitability for the description of complex human motion. We have derived an optimization heuristic, which is used to automatically generate the initial motion space. We have evaluated our system by comparison against ground-truth motion data and alternative methods.