人体运动建模的实时逆运动学算法

A Real-Time Inverse Kinematics Algorithm for Human Motion Modeling

人体运动的虚拟仿真及建模是当今计算机图形学研究的一个热点,而逆向运动学方法则是求解人体运动的一项关键技术.为了实现人体运动建模,提出一种新颖的基于人体关节约束的实时逆向运动学算法.首先基于骨骼长度改进了传统的循环坐标下降(CCD)算法,使其生成的运动效果更加流畅自然;然后引入生物分子运动模拟中的SHAKE算法,使短距离运动建模效果更加高效、鲁棒.在此基础上,提出了模拟多个节点同时运动的解析算法,以获得在多个关节点共同作用下的人体运动仿真效果.同运动捕捉的实测数据相比,采用文中算法所得的结果与在视觉效果上非常接近,并通过实验证实了其应用价值.

In recent years, virtual human motion modeling has been a hot research topic in computer graphics. Inverse kinematics is a key technology for motion modeling. A novel inverse kinematics algorithm based on the human constrained joints for real-time human motion modeling is proposed. We improve the traditional cyclic coordinate descent （CCD） algorithm based on bone distance to make the motion more fluent and natural. Then, we incorporate SHAKE algorithm, which is normally used in biological molecules motion simulation with the CCD algorithm for short-distance motion modeling. Furthermore, we develop an analytic algorithm accounting for the concurrent the motion of multiple nodes, to facilitate the simulation of human motion effected by multi-joints. The result of our algorithm is very close to those generated with motion capture data, which indicates the application potential of our algorithm.