A comparison of motion capture data recorded from a Vicon system and a Microsoft Kinect sensor

This experiment investigates the perceived differences in the quality of animation generated using motion capture data using a Vicon motion capture system and a Microsoft Kinect sensor. The Kinect uses a depth map camera to determine the position of users and objects within the view of the camera. Using this information, the depth map can be processed to track the motion of a user over time, and for this experiment, this was done with PrimeSense's NITE algorithm[PrimeSense]. The final recording generated from the depth map camera is noisy and is clearly a lower quality recording when compared to the motion capture data created using a Vicon system. To overcome this problem, the motion capture data from the Kinect can be filtered to smooth the noise and create more usable data. This experiment focuses on using a Butterworth filter to smooth the Kinect data and attempts to identify how much the data can be smoothed so that the quality of the motion capture data is indistinguishable from the Vicon data for short video segments.