2P1-N06 Attitude Representation for Motion Feature Extraction using Singular Value Decomposition(Digital Human)
Takafumi OKAMOTO, Hiroaki Nakanishi, Sayaka Kanata, Tetsuo Sawaragi, Yukio Horiguchi · The Proceedings of JSME annual Conference on Robotics and Mechatronics (Robomec) · 2012
Singular Value Decomposition (SVD) plays an important role in analyzing human behavior. Human behavior can be represented using attitudes of human body parts. Unit quaternion is widely used to represent attitudes because of lacking singularity. In the previous study, unit quaternion was used to represent attitudes of human body parts. However, a constraint exists in the representation using unit quaternion, and it has significant effect on the feature extracted by SVD. The purpose of this paper is to establish a proper representation of attitudes of human body parts for behavior analysis using SVD. Behavior analysis results using unit quaternion were compared with the results using exponential map, and the contribution ratio of the first mode in using unit quaternion is too large to analyze human behaviors properly. Finally, we concluded the exponential map is suitable attitude representation.