About filtering and rejecting invalid analog eye-data samples

Hi Jaewon, ML users,
I am taking in analog gaze x,y input and hoping to use it for calibration. Turns out our iscan eyetracker sends 0,0 when it is not able to track the eye. This creates problems for gaze contingent experiments where we center the origin when the monkey is looking at the desired stimuli and then want to enforce say a 2-3 dva gaze box. Even if the monkey is looking correctly, the trial breaks when the tracker loses the eye and sends 0,0.
Is there a matlab script / hook that I can attach to the routine that reads in analog x,y (and eventually shows it as a red dot on the visualisation) so that I can reject such 0,0 cases as we read in analog data. An exact 0,0 is very unlikely as a valid eye position in my case.
Advise on this will be great.

best,
Harish

You may want to check your AI ground configuration. Sometimes misconfigured ground causes a similar problem.
https://monkeylogic.nimh.nih.gov/docs_NIMultifunctionIODevice.html#AIGroundConfiguration

If you are 100% sure that it is a problem of your eye tracker, see the following post. You can do linear interpolation after removing [0 0] in your custom function or replace [0 0] with the last xy coordinates that occurred before the [0 0] (or the first xy after the [0 0]).
https://monkeylogic.nimh.nih.gov/board/read.php?3,32

Hi Jaewon,
Thanks for your reply. Yes I am sure that the tracker sends out 0,0 when it doesn't find the eye. I see that the example code modifies a joystick xy to say xy+[a,b] . In my case it would be nice to have a way to make such xy values NaN and if ML can then ignore such points, is this possible ? Since a [0,0] in our case means the tracker doesn't know where the eye is, do you have any thoughts on this?

best,
Harish

I already gave you a solution but it seems that the message was not clear enough. I rewrote my previous post. Please read it again.