Research On The Technologies Of Virtual Human Motion Editing And Synthesis Based On Data-driven

Virtual human face current difficulties in complex work, bad motion controllability and low data reuse rate, we put forward the technologies of virtual human motion editing and synthesis based on data-driven, which mainly based on the existing human motion data to generate new motion. In this paper we mainly focus on the following aspects:First, we proposed a new extraction method based on wavelet transform key frame aim at solving the problems that traditional quaternion method need a large amount of calculation ,that distance between quaternions are difficult to measure and that the curve simplification method of key frame extraction lost critical gesture. By selecting a different threshold value, this method realise the key frame extraction from motion sequences. We can achieve different gesture movement by editing the key frames. It not only maintains the original motion characteristics, but also could achieve movement to meet other requirements.Secondly, we put forward motion retargeting method based on the different topological structure, in view of unreality of virtual human's motion after retargeting and its reorientation limitation only to the same topology structure of the human body. This mothod reduce the degree of complexity of redirection algorithm through adjusting human skeleton model of different topological structures, and ensure human motion more realistic and natural after retargeting.Finally, there are many methods for motion synthesis, but its function is very simply. In order to solve this problem, we design the virtual human motion synthesis system through Poser software and combine with Poser and Matlab software to generate motion synthesis of virtual human motion effect and verify the validity of the method of virtual human synthesis.Through these simulation experiments, the technique of data-driven for virtual human motion editing and synthesis can generate reality and complex human motion efficiency.