Research And Implementation Of Motion Synthesis

With the development of movies and games industry, motion synthesis has become a hot research point these years. Motion synthesis focuses on how to use existing motion clips to generate new motions. Motion Graph, which is considered as an efficient method of human motion synthesis, has been widely used.Motion Graph uses a graph structure to manage and connect different motion clips. General method of Motion Graph construction consists of normal motions, like walk, jump and run, seldom of them includes motions of multiple styles, for example, run happily and walk proudly. If a Motion Graph consists many motions with different styles and these motions could transited to other motion naturally, then we could make more realistic human motions according to emotion change or the requirement of stories. Motion synthesis with multiple styles is a better choice for 3D animation and game design. Thus it has great practical value.However, this is not easy to achieve. Adding new motion styles to the original Motion Graph will cause huge increase of the construction time and search time, due to the fast growth of Motion Graph size.In order to solve this problem, we proposed a method to create a modified Motion Graph with structure table. Structure table is a new concept in this paper, which could store the position information of each motion clips in the Motion Graph. This information helps to create a better and more efficient Motion Graph. Local optimization is also allowed during the construction. By introducing structure table, search of Motion Graph has been turned into a table look-up process or local search. Efficiency of construction and search of Motion Graph are highly increased in our method. In addition, when there is no direct link between different motions, we introduce a new structure called navigation map. By using navigation map, we could link two un-directly connected motions via a series transition motion clips.Experiment results show that by using structure table, we could largely reduce the construction and search time of Motion Graph. In addition, connectivity is guaranteed and many unreasonable motions are avoided. Finally we make it possible to create Motion Graphs with multiple motion styles. Navigation map raise the possibility of successful search and meet the customers'requirement.