Motion capture data can be characterized as a series of multidimensional spatio-temporal data, which is recorded by tracking the number of key points in space over time with a 3-dimensional representation. Such complex characteristics make the processing of motion capture data a non-trivial task. Hence, techniques that can provide an approximated, less complicated representation of such data are highly desirable. In this paper, we propose a novel technique that uses temporal clustering to generate an approximate representation of motion capture data. First, we segment the motion in the time domain with an optimal partition algorithm so that the within-segment sum of squared error (WSSSE) is minimized. Then, we represent the motion capture data as the averages taken over all the segments, resulting in a representation of much lower complexity. Experimental results suggest that comparing with the compared methods, our proposed representation technique can better approximate the motion capture data.
|Title of host publication||Proceedings of the 15th ACM SIGGRAPH Conference on Virtual-Reality Continuum and Its Applications in Industry - VRCAI '16|
|Place of Publication||New York|
|Publication status||Published - Dec 2016|