Segmentation of motion capture data based on measured MDS and improved oblique space distance

Abstract

The segmentation of motion capture data is essential for the synthesis of motion data, its purpose is to split long movement sequence data into many different independent semantic motion clips, and it requires that the segmentation of motion capture data is effective and accurate. This paper proposed a segmentation algorithm of motion capture data based on measured MDS and improved oblique space distance. The proposed approach used the multidimensional scaling (MDS) to achieve the space mapping from original high-dimensional data to low-dimensional, and then calculated the improved oblique space distance between frames in the specified windows and the preceding section in the low-dimensional space, and obtained the final segmentation points by similarity detection. Finally we obtained the independent semantic motion clips, and we verified the feasibility of the algorithm through experiments, and the accuracy rate of our method is improved compared with the traditional algorithm.