Three-Dimensional Virtual Human Motion Synthesis Technology Research Based On The Skeleton

Practical activities of human society depend on the participation of people in the final analysis, the study of human complex behavior characteristics has important realistic significance for the development of human society. With the rapid development of modern computer science, the use of virtual human (or avatar) to simulate the behavior of complex human nature has been become the research focus. Virtual human motion synthesis technology is one of the important research content in human motion simulation, involving in computer graphics, kinesiology and so on. It has a series of multi-dimensional, multi-level, multi-disciplinary, and multi-objective characteristics.The research goal of this thesis is to present a kind of three-dimensional virtual human motion synthesis technology solution based on skeleton, building the skeletal model of virtual human, and generating real and natural diversification virtual human motion sequences. Firstly, on the basis of analyzing the hierarchical structure of the human, we obtain the human model data and simplify the process of model data based on the principle of human anatomy to build the virtual human skeletal model, which is fully compliant with the H-Anim standard. Then, we select the key frames, classify key frames simply, and represent the key frames with vector and rotation quaternion. During the process of key frame interpolation, the general linear interpolation is used for virtual human space position changes, the changing process of motion posture is more complex that need quaternion spherical linear interpolation algorithm. If adjacent key frames of data is abnormal in the interpolation process, it cause the interpolation result is not correct. In order to make up for the lack of key frame interpolation method, we employ motion constraints reconstruction to adjust the motion posture, which synthesizes the real and natural motion sequences. The state of virtual human motion is changed in real-time.In order to meet the needs of real-time changes and improve reusability of motion data, we select two different style motion sequences to synthesize a new state of motion. The two inputting motion sequences are warped in the time and space, so that the key frames of the two motion sequences generate corresponding relationship to synthesize the real and nature diversification motion sequences.