Abstract :

Motion retrieval in the designer's way is always preferred by the animators in motion reuse for character animation, therefore how to naturally draft the desired motion in the animator's mind and effectively convey it to the computer becomes a crucial point in content-based motion retrieval. In this paper, we present an intuitive motion retrieval approach in which the query motion clip is acquired by asking the animators to roughly perform them together with motion sensors. This overall retrieval process is composed of three major steps. The animator firstly places the specific sensors on his body, and then the computer creates the mapping between the joints in 3D skeleton and the motion sensors by posture alignment. Secondly, the animator performs the desirable motion in his mind, and the system automatically converts the mapped movements into the query motion clip. At last, the system looks for the matched motion clips from the motion capture database. We present experimental results on a test data set of more than 230 minutes of motion data. It shows that our method not only provides the visual communication among animators, and but also greatly improves the retrieval effectiveness in motion reuse.

Keywords :

Content-based motion retrieval - Motion sensors - Perceptual user interface

Paper :
Computer Graphics International (CGI) 2008 Conference - contact first author for the PDF file.

BibTeX :

@inproceedings{LZLPGM08,
author = {X. Liang and S. Zhang and Q. Li and N. Pronost and W. Geng and F. Multon},
title = {Intuitive motion retrieval with motion sensors},
booktitle = {Proceedings of Computer Graphics International (CGI), Istambul - Turkey},
year = {2008},
month = jun
}