Automatic subspace synthesis of motion styles based on isomap

Abstract

Motion synthesis methods based on 3D motion data are extensively studied in recent years. And 3D motion styles synthesis has been rarely explored. In this paper, a method for automatic and quantitative synthesis human motion styles is proposed. First extended ISOMAP theory is used to map original motions into low-dimensionality subspaces, which can reserve the intrinsic properties of original data. Synthesis is applied in such new subspaces and motion styles can be reconstructed. Experimental results show that our methods are effective. © 2008 Springer-Verlag Berlin Heidelberg.