Share Email Print

Proceedings Paper

Similarity retrieval of motion capture data as time-series
Author(s): Jiale Wang; Yuanjun He; Haishan Tian
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

This paper proposes a new method for motion capture data retrieval. To measure the similarity between motions, we define two distance functions: the local distance function and the global distance function. The local distance function is to measure the similarity between stationary body poses, and it is based on the weighted position distance between joint-pairs. We take the particularity of end-effectors into account by assigning them greater weights. The global distance function is to measure the overall similarity between motions. Because motions are time-series, it is necessary to align them on time axis to make the logically corresponding events at the same time point. We use the timewarp curve to describe the corresponding relationship between frames of two motions, and use the dynamic timewarping algorithm to find out the minimum sum of the local distances between all corresponding frame-pairs. This minimum sum is namely the global distance that measures the overall similarity between motions. The experiment demonstrates the effectiveness and accuracy of this method.

Paper Details

Date Published: 2 December 2005
PDF: 7 pages
Proc. SPIE 6045, MIPPR 2005: Geospatial Information, Data Mining, and Applications, 60450R (2 December 2005); doi: 10.1117/12.650378
Show Author Affiliations
Jiale Wang, Shanghai Jiaotong Univ. (China)
Yuanjun He, Shanghai Jiaotong Univ. (China)
Haishan Tian, Shanghai Jiaotong Univ. (China)

Published in SPIE Proceedings Vol. 6045:
MIPPR 2005: Geospatial Information, Data Mining, and Applications
Jianya Gong; Qing Zhu; Yaolin Liu; Shuliang Wang, Editor(s)

© SPIE. Terms of Use
Back to Top