Teach and Playback of Realistic Target Reaching Movement of Virtual Human Upper Limb

Abstract

The significant research objective of virtual human is to generate realistic movement like real-life person. For the most frequent daily activity of target reaching movement of human upper limb, teach and playback method is proposed to generate vivid human motion, which combines respective merits of parametric keyframe, kinematics, dynamics, and motion capture methods together. First, the tip trajectory and velocity rules are learned and estimated by a weighted least squares support vector machine (WLS-SVM) and a combined half-Gaussians model respectively. Then an inverse kinematic method with dynamic constraint, damped least squares (DLS), is used to calculate the joint angles of the upper limb. Finally, the experimental results show that once the target position is appointed, the 3D virtual human skeleton model can generate realistic reaching movement.