Origin And Gain Calibration Method

--------------------------------------------------
Dear Jaewon, sorry for the naive question, butI I'm a beginner in this field, and I would like to figure out how does the origin and gain method work.
My set up is based on the ViewPoint EyeTracker(Arrington research). What I get is in fact the pupil position(in X-Y coordinates) on the screen in respect to a point(the center). So in the moment I set origin(when the subject is looking the center and the reward is released) I get the [0 0] point of regard(I mean the postion of the pupil at the center of the screen mapped in x-y coordinated). Then, I present a peripheral target, let's say bottom right. Since I experience an underestimation of the actual saccade I increased the gain. Then I present the same target until I can't see the signal (and ear the reward delivery)on it. At this point my calibration should be done. But sometimes, after adjusting the gain,I present the center again and I loose the center(I mean according to the fact that the gain is increased if a present the central target the signal is not reaching it anymore). Should I re-set the origin after each gain adjustment? How effectively the method works? I read about POR coordinates estimation based on 5 to 9 point of calibration and I wonder how the origin and gain can require just one point. Furthermore, in the GUI to what parameters correspond "rotation"?. Many thanks in advance
Best
And

It seems that you did the procedure all right. Yes, your calibration should be done when you finish adjusting gains, but you may find the origin moved a little, if the initial gain was small and therefore did not provide enough resolution for you to set the origin precisely. Then you can just repeat the procedure a couple of times. As the gain increases, you will see better results each time.

If the origin is not stable at all and moves away from its initial position to different directions every time, check the voltage output range of the eye tracker. The eye tracker has its own output gain and you may see unstable results because the eye tracker's gain is small.

The ViewPoint Eye Tracker has a simulation mode so that you can mimic eye positions with mouse. It is easier to find problems with the simulation mode. Try it.

The "rotation" turns the directions of XY axes. It is useful when your camera is not level.

Many thanks

Dear Jaewon,
I experienced the same problem of eye calibration with origin and gain method. My Eye-Tracker system is the Arrington Viewpoint Eyetracker. As you can see in the picture attached below, the camera is positioned on the top, at the center of the monitor. Today I was recording the signal from the right eye. As first, before setting origin and space, I saw the signal in the left part of the calibration window (i.e. the monitor where I present the stimuli). So, instead of presenting the central target, I firstly moved the offset to the right. Then I presented the center and the value of the origin was [x=-4 y=2]. Note that these were the only values through which the monkey was able to fix the central target. According to my understanding, these values mean that when the eye is fixing the center the voltage recorded at this point is [-4 2]. Is this correct?. Then, I adjusted the gain for more peripheral saccades and I saved the values. But when I run my eye-task I see the signal completly shifted to right. I then re-calibrate with less negative values in X (let's say -2) and I lost the right part. What does this means? How can I adjust the gain in the Eyetracker? I went through the user's guide but i didn't find anything about that. How can I test the output voltage that ML receives from the Eyetracker? (Note that the connection between the two system is via ethernet-cable).

I am not sure how you ran the calibration. Especially I don't understand these two parts: "instead of presenting the central target, I firstly moved the offset to the right" and "I then re-calibrate with less negative values in X (let's say -2)". They do not make any sense to me. Please read the manual and follow the instructions there.
https://monkeylogic.nimh.nih.gov/docs_CalibratingEyeJoy.html#OriginGain

The first thing you need to do is to present the central fixation point (by clicking the "Show Center" button) and press the space bar when the subject looks at it. Then adjust X & Y gains so that the eye tracer can reach peripheral fixation points you click when the subject looks at them.

The calibration used during the task is exactly the same as what you see in the calibration tools. If you don't see the eye tracer moves onto fixation points you click in the calibration tools, your calibration is already wrong. Don't be bothered to run a task to confirm it.

Dear Jaewon sorry for the bad explanation, but the point is that I start the calibration procedure with raw-signal. How do I know that the monkey is looking at the center? If the signal is uncalibrated and appears only in the left top part of the window (i.e.monitor), the signal never enter the central target window even if the monkey is looking at it. So when should I set origin and space?
Best

Raw-signal means no calibration. Choose one of the other options.

The signal never entered the central target window, because you did not set the origin. Once you press the space bar, it will come to the center of your screen.

If you are not sure where the subject is looking at, just use your best guess. If the result is not satisfactory, try again multiple times. If the subject is not trained for fixation, train it first. Ask your supervisor and colleagues what to do.

And stop calling me Jeamon, please.