The meta problem you are writing about is...How given many people in the Kinect's view can I select only 2 users?

Much of that is your choice based on the desired design/look/functionality of your patch. Is it the 2 closest users? Is it the 2 users jumping up and down? Is it the 2 users waving their hands? What happens if a 3rd person walks in front of person #2? Do we ignore that new person? ...and so on...

NITE, the technology that does the user/skeleton tracking has its own internal algorithm to detect and track the skeletons. There was never a working method to define for NITE what skeletons to track. So...NITE tracks anything it can see and your CPU can process. There was never a published API to determine the max number of users that NITE could track. So I coded jit.openni to support up to 15 (I never had more than 3 reliably tracked).

Therefore, it comes to your code to decide/choose which users to track. You will need to think through questions like I wrote above. After thinking on these, then you can write your patcher with that logic and use the userid values to filter joint/position data. It is possible that NITE may be tracking users that you never use in your patch.

Thanks a lot for this answer. I was wondering if was possible to say to the jit.openni just track 2 (as you coded it for track 15). When this is not possible then of course I will do it in the patch with logical things :) Thanks again a lot!