If my understanding of the device is marginally accurate, I understand it offloads almost all of the processing to the 360. So, I would imagine the raw data coming out of it is like getting raw input from 64 mice at the same time. If anything is done directly on the device, it is probably "stabilization", like muting the background and focusing on a central moving object, but actual tracking and number crunching is done off of it.

Here's a video of a guy looking at his room with IR goggles. This observation makes me think that there is a Lot of data to be crunched to get an accurate use of it. http://www.youtube.com/watch?v=r7nRKU0nFxA