I want to drive a human model with each joint's position only.And these joints' position is not in high accuracy.It's means that each bone's length is not fixed from the joints' position data.But the human model's bones' length are all fixed.

I've found that the Kinect SDK v1.8 would give each tracked joint's position and hierarchical rotation.I think the Kinect's tracking result for joint position isn't in high accuracy,too.And i think the joints' hierarchical rotation are calculated from joints' position.The XBox game shows that the Kinect's hierarchical rotation get a good result to drive a human model.

So,i want to ask how to get each joint's rotation with its position only?

Any idea is appreciated.Thanks a lot!