Research on Common Structure of Motion Data

The movement process of organisms contains a large amount of information, which leads to different structure of motion data in different applications. How to construct a data structure to contain more common information of motion is a topic that researchers have been exploring. It is essential to set a series of criteria for the motion data structure to judge whether it can be a good representation of campaign information and fast and efficient readed/writen and calculated by computers before construct it. In this paper, we proposes 12 construction criteria for motion data structures through analysing the main current application scenarios of motion data, the way computer process data and the requirements of deep learning on data. We design a common structure of motion data CSMD based on these 12 criteria, which can be adapted to almost all common application. Our method is 78% faster than traditional methods. We also build a toolkit to process common motion data including C3D, ASF/AMC, BVH, CSMD, etc.