三维人体运动特征可视化与交互式运动分割

Feature Visualization and Interactive Segmentation of 3D Human Motion

为了从运动序列中提取不同类型的运动片段进行人体动画创作,提出一种三维人体运动数据可视化与交互式分割技术.首先,采用人体各主要骨骼夹角作为对原始运动数据的几何特征表示,并提出一种启发式方法自动检测其潜在分割点,最后将提取的运动特征可视化并使用可交互用户界面对其进行精确分割.实验结果表明。该特征表示及交互式分割方法能够更加方便、准确、高效地对包含多种类型的长序列三维人体运动数据进行分割.

In order to get primitive actions for animation production, this paper proposes a visualization and interactive segmentation method for 3D human motion sequences. Firstly, a new feature based on bone angles of human skeleton is used to represent the original motion data. Then a heuristic method is introduced to detect the candidate segmentation points on the feature sequence. Finally, a graphical user interface is created to visualize the candidate segmentation points and pick the precise segmentation points interactively. Experimental results show that this method can divide motion sequence into primitive actions more conveniently, efficiently and precisely.