Research On Motion Retageting

Recently, motion capture has become one of the most promising technologies in character animation. Realistic motion data can be captured by recording the movement of a real actor with a motion capture system, and motion retargeting will adapt these motion data to new character. While the target character is different from the original one, the target character is likely to lose desire features of original motion. The problem can be solved through motion retargeting technology. This thesis introduces the standard of the virtual human model, and builds our virtual human model firstly. Then we show the hierarchical structure of the model and the distribution of the joints. We also analyze the importance of motion capture device and focus on how to capture sign language data through data glove. With the assistance of data glove we propose a new method for retargeting sign language data to different virtual human model with different sizes and proportions. Realistic and natural animations can be produced to express similar meanings to the original. The proposed method first defines many sensitive points on the human body and selects the key sensitive points through analyzing the importance of the sensitive points. Next a novel mapping method based on relative position is presented to adapt the original sensitive points to the target sensitive points. Finally we utilize an IK solver to realize the retargeting problem. Our method had been applied in the new version of china sign language electronic dictionary. Experimental results show that the proposed method dramatically improves the recognition rate about 30%.