AUTOMATIC KEY-FRAMES EXTRACTION OF HUMANOID MOTIONS

Chin-Hung ； Ko Jia-Yi Li ； Tain-Chi Lu

《技術學刊》 32卷1期 (2017/03) Pp. 39-47

摘要

3D authoring tools provide a wide variety of motion editing functions to users to make motions appear more plausible and natural. Animators obviously need to spend much time and invest considerable efforts to edit great amounts of frames for the sake of obtaining their desired animation results. Therefore, animators try to reduce their burdens by picking up the appropriate key-frames, editing these chosen key-frames, and interpolating them for their desired results. Even animators with sufficient experience in animation may need to use several number of times to gain the appropriate key-frames. In this paper, we present a method to overcome aforementioned key-frame problem by automatically picking up the most significant keyframes from a continuous motion clip. Users only need to input a target motion clip in Biovision Hierarchy (BVH) format and provide a series of parameters, and the proposed method will extract the most significant keyframes from target motion clips automatically. In order to verify the quality of extracted key-frames, we directly interpolate the extracted key-frames and compare them with the input motion clip. The experimental results suggest that the proposed method is less labor-exhaustive, and it decreases the operation time without sacrificing original motion features.

關鍵字

character animation ； principal component analysis ； motion saliency ； key-frame extraction