Research And Implementation Of Motion Synthesis Based On GPU

In computer graphics field, computer simulated3D animation hasbecome a hot area of research. Computer animations, especially thosebased on motion capture are now fully used in movies and games. Thetechnique of reusing motion captured data is called motion synthesis.Key-frame is the supporting technique of motion synthesis, it allowsuser to assign key postures of a character in animation, and computer willthen take the rest work, which is interpolation. If we want computer togenerate new motions by need automatically, we have to use motiongraph technique. But common algorithms on motion synthesis havedisadvantages on speed and are resource-intensive. As the fidelity level ofthe animation increases quickly, the motion graph data, skin dataexplodes. A new mind of resolving computing ability is needed urgently.The fast growing of GPU parallel accelerating technique recentlyprovides a better way to solve massive data problem.In this paper, we implemented and improved motion graph algorithm,by constructing a database of thousands small motion clips, we generatenew motions. Furthermore, we use skinning to enrich animationcharacters to maintain higher fidelity. We introduced new algorithm basedon GPU parallel accelerating into motion editing, motion graph andskinning. These algorithms are implemented and run on GPU, whichlargely increase the speed and reduce resource occupation. Experiment results show that by our improved algorithms using GPU,the real-time, automatically generated, high quality animation has becometrue. Using our method, the efficiency of animation production increases,it can satisfy more user requirement.