Motion Capture Data

YART provides a motion capture (MoCaP) parser in a bvh format. To test the functionality, we parse CMU motion capture (MoCap) data in a bvh format and form kinematic chains that are used to define the rigid body of a robot. CMU MoCap data can be downloaded in this github.

A bvh format naturally provides a kinematic chain with revolute joints. So it is straightforward to import the skeleton structure parsed from a bvh file into a kinematic chain. A minor difference is that whereas the rotation of the kinematic chain of a robot is defined by a rotational axis (a) and the angle (q), the rotation of the joint of a skeleton is simply a spherical joint and defined using an Euler angle. Note that we also rotate the skeletons so that the z-axis (hip to head) is heading up (Z-UP), and the root joint (hip) is centered at the origin, to match with our convention.

The image below shows the kinematic structure of the skeleton of the CMU MoCab DB.