Improved kinematics based motion compression for human figure animation

This paper presents an improved kinematics based motion compression algorithm by using key frame extraction. In a motion chain, the movement of each joint is represented by a series of curves in 3D space. In general, when compressing the motion data, the distortion of parent joint caused by quantization in turn affects its child joint and is accumulated to the end effector. The inverse kinematics based method can address this problem well. Moreover, to achieve a higher compression rate, we propose a period detection method and a key-frame extraction method into 3D position curves of end effectors. Our method automatically detects the key frames in motion curves of end effectors using frame decimation. Combining inverse kinematics technique and key frame extraction method, our method achieves efficient compression with a high compression rate and high quality of the motion data. By comparing with a conventional prediction we demonstrate the advantage of ours with some examples.