Motion Analysis and Application Based on Wavelet Transformation
LIU Feng, ZHUANG Yue-ting, PAN Yun-he, LUO Zhong-xiang (Department of Computer Science and Engineering, Zhejiang University, Microsoft Visual Perception Laboratory of Zhejiang University, Hangzhou 310027)
Motion editing is the key technology to improve the re-use of motion captured data and produce complex human animation. However, few techniques are capable of editing motion at high level. In this paper, wavelet transformation is introduced into motion multi-resolution analysis and some new algorithms, namely motion feature enhancement, motion fusion and motion feature extraction and synthesis, are proposed. Motion signal is decomposed into multi-resolution levels with wavelet analysis. The coarse level represents the overall pattern of a motion signal while the fine levels describe the details. Special motion style can be highlighted through enhancing the corresponding level content. And multiple motions can be fused by multi-resolution blending to create new motions somehow like to the blended motions. Special motion style can be synthesized into other motions by texturing them with related fine levels extracted from the related motions. The experiment shows that these algorithms are well suited for motion feature editing and enable animators to edit motion effectively at high level.