Research On The Key Techniques For Virtual Human Based Motion Capture Data

This paper studied the virtual human based on motion capture the latest key technology problems-based on motion capture virtual human motion character extraction, based on motion capture virtual human movement recognition, based on inverse kinematics virtual human control and virtual human movement path to edit. The main research contents of the topics are as follows:1. Using nonlinear KPCA constructed to segmentation algorithm of comprehensive features of function and the corresponding evaluation function; Determine the function optimization nuclear, then applies get formula of kernel function of virtual human movement characteristics formula extraction.2. The decision tree machine learning, more examples learning and hidden markov models based on motion capture data into the virtual human movement recognition. To identify common motion model than the original algorithm, and high recognition rate.3. Introduced based on forward kinematics, inverse kinematics method virtual human control method. Based on analytical proposed the inverse kinematics virtual human control algorithm is given, and the formula of inverse kinematical constraints.4. The use of based on constraint cubic spline B virtual person trajectory of editing, generate new sports method.