Smoothing Motion Capture Data

As part of my uni course I have to create a piece of original research into something in the CG field. For my project I designed and implemented a novel method of skeleton data smoothing. (This is the project I talked about a couple of posts ago).

I’d like to share some of the results of my research in a couple of videos. First video shows tracking of star jumps, to show the much smoother animation of my system compared to raw tracking data:

My system takes the tracking data broadcast across the network at 20Hz and plays it back at 60fps using a Kalman filter to interpolate across frames.

 

The Second video shows the estimation/fitting method I developed, applied to a couple of stretches. You should be able to see that the system makes accurate predictions about where the limbs should be in some cases or returns to a rest position if we have no data.

I’m really happy with the results we have achieved for the project and have submitted it to a conference for a poster submission and hope to get it accepted!

Leave a Reply

Your email address will not be published. Required fields are marked *