Pose Synthesis of Virtual Character Based on Statistical Learning

We present an inverse kinematics implementation technique based on statistical learning. Because of the high dimension of character animation motion data, direct analysis on them is a very hard work. We map the motion data from high-dimensional observing space to two-dimensional latent space, based on Gaussian process latent variable models (GP-LVM), then, find out the representative poses of virtual character by clustering the motion data in latent space. Finally, weight the representative poses and optimize the weights, combined with constraints on the end effectors, and synthesize the optimized pose. The experiments show that our method obtains satisfying effect.