Low Level Segmentation of Motion Capture Data based on Hierarchical Clustering with Cosine Distance

Abstract

D motion capture is to track and record human movements. In recent years, it has been applied into many fields, such as human computer interaction, animation, etc. Low- level segmentation of motion capture data is of significance to the various applications of 3D motion capture; however, due to the high dimensionality of motion capture data, traditional low-level segmentation methods can hardly work out a suitable segmentation for motion capture data. In order to solve this problem, a low-level temporal segmentation algorithm based on cosine distance is proposed, hierarchical clustering is explored so that similar velocity vectors are clustered together according to the cosine distance in a progressive way, the center of each cluster is updated as the vector derived with linear regression, the segment boundaries are determined as the point when the cosine distance between adjacent velocity vectors is greater than 1 (angle>90 degrees). We have conducted experiments on the motion capture database provided by Carnegie Mellon University (CMU), the experiment results show that the performance of the proposed method is optimistic.