Vizard and Polhemus Patriot
 

We are having a problem with the calibration of our Polhemus patriot tracking device running through Vizard as per 'SekulerExperiment.py' that you wrote for us. We started PiMGR, set up the hemisphere tracking, and closed it again, and run the experiment file in Vizard. We keep the orientation sensor in the same place every time, and calibrate with the glove in the same position. Yet every time we do that, the glove starts in a different position on the screen. This is with the same subject in the same location, same arm length and position, and everything. Any clues as to why this happens? Has anyone else experienced this in their testing?

So with each run of the Experiment the hand is in a different position? This is without going back into PiMGR?

That's correct. PiMGR is closed at the time. We open Vizard, run the experiment (a couple of trials), close the experiment window, and run the experiment again. Both runs have the hand in a different position. Note that this also happens if we leave PiMGR open when we run the experiments.

How different is the position? Does it feel like some axis are reversed when you move the hand around?

No, just that the hand moves in XYZ without any rotation if we calibrate the same way. Occasionally there will be a little rotation, but that is rare. However, we can try solutions for both things (rotation and nonrotation orientation errors).

I am confused about what the problem is.

1)Is the movement of the hand incorrect? For example, does the hand not rotate when you move your hand in the real world?

2)Is the positioning of the virtual hand, reletive to the viewpoint, change with every run of the script?

The movement of the hand is incorrect.

The second is correct. The position of the virtual hand, relative to the viewport, changes with every run of the script, even if we keep all parameters the same (keep the hand in the same initial position, the orientation cube in the same position, etc.).

Ignore previous post
 

I apologize, I'm trying to edit the previous post (the answers to the 2 questions), but cannot. Here's the real answers:

The movement of the hand is correct. The hand moves as I move my hand in the real world.

The second is problem that is occurring. The position of the virtual hand, relative to the viewport, changes with every run of the script, even if we keep all parameters the same (keep the hand in the same initial position, the orientation cube in the same position, etc.).

I ran the SekulerExperiment here and found that if I left the sensors on the ground in the same position the hand/arm position did not change with each run of the script.

This suggests that something is wrong with the shoulder calibration (though there could be other sources of this behavior I am not thinking of). Try printing out the raw sensor data from the shoulder calibration sensor when the calibration occurs and see if this significantly different between script runs.