Research On Motion Synthesis Method Based On Motion Graph

The research on virtual human motion synthesis technique is one of the most difficult and active research areas in virtual reality,and it is also one of the core techniques in the cultural industry.It has broad promising applications in many areas such as movie and video animation,3D games,preview of security,etc.Motion fidelity and control flexibility are required in these applications.In order to meet application needs of high realistic and interactive virtual human motion,this thesis studies on motion analysis and synthesis method based on motion graph,it consists of three steps:data preparation,construction of motion graph and its application.The contributions of this thesis are as follows:1.An automatic and robust segmentation of human motion capture data method based on motion features is proposed.Reasonable segmentation of human motion capture data is prerequisite and foundation for motion graph construction.In the process of motion synthesis based on motion capture data,the shorter the length of motion data is,the more flexible control motion synthesis can get,and in order to control human motion,the data reused for motion synthesis should be long enough to contain meaningful semantics.We extract features based on feet relationship.Through matching the motion capture data to motion template based on the motion features,we implement automatic segmentation, avoiding fussy manual operations and improving the efficiency and accuracy of the motion data segmentation;at the same time,this method is robust,segmentation correctness is enhanced by 34.12%compared with the widely used segmentation method proposed in[KS05].2.Motion graph method based on parametric motion synthesis is proposed.Motion graph method is an effective method of data organization,motion graph is a directed graph composed of nodes and edges.In existing methods,motion capture data is directly stored in motion graphs,which limits the capacity of the synthesized motion and the user's control accuracy.We propose motion graph method based on parametric motion synthesis,node in motion graph stores continuous control parameter space,given a specific parameter such as motion velocity,turning angle, etc,according motion segment can be synthesized accurately,which enhances the description capacity of the motion graph and user's control accuracy.Edge in motion graph denotes transformation relationship between different parameter spaces,smooth motion transition between motion segments can be get through motion blending, meanwhile,root position is computed after posture blending,which avoids foot sliding and root orientation jitter in existing blending methods,and enhances fidelity of human motion. 3.For specific application,path-based parameter extraction method is proposed.In the above motion graph,the input is specific parameters of physical meaning such as motion velocity,turning angle,etc.However,it is required to provide user with a flexible interactive mode to control virtual human motion,only path trajectory, for example,and on-line motion sequence along the trajectory is required to synthesize.So it is required to extract specific parameter from intuitive path trajectory.In this thesis,a path-based parameter extraction method is proposed,which not only satisfies the on-line requirement,but also makes the synthesized motion consistent with the requirements of path trajectory by closely combining path trajectory with motion data attributes.The experiment results show that this method can get accurate on-line motion sequence in accordance with the requirement of path trajectory.