Calibration - A line instead of points?

Place to discuss eye tracking and related topics

Calibration - A line instead of points?

Postby jeffrey » 30 Mar 2016, 17:38

Why do we calibrate points, and not lines?

I would assume that the human eye is way better in tracking a moving object, then a static point and thus would result in a better calibration.
I do understand that it is easier to start calibrationRegistration, provide point, wait a little, stop calibrationRegistration and repeat that for 9 points. It is also easier to recalibrate a single "point" this way.

But I would assume that its not that difficult to feed the current x/y position of the point live to the API while it is moving, to use it in combination with the current gaze data to calibrate. If the screen resolution is known to the calibration script, it can proved feedback about what area's are not yet calibrated correctly, or provide the point it would like the user to look again. The point can than move towards those area's/points to recalibrate those area's.

Instead of correlating points with another, the calibration script can correlate figures with another. That would (I assume) reduce the noise in the calibration process.


I have ran in multiple problems using the V1 eye tracker.
- Some (ill) people are unable to calibrate 9 points, it takes too long. For some of them, 2 calibration points (left/right) would suffice! (just to speak yes/no)
- People with a glass eye are unable to calibrate. Yet, when the calibration has been completed, the system can work with one eye only.

I thus would be eager to know if:
- In the future it would be possible to access the raw data: The camera data
- In the future it would be possible to access the raw data: The x/y position + eye angle (vector) of both individual eye's
Using that, it would be possible to create your own calibration procedure, e.g. the one I suggest above.
- It is or will be possible to calibrate the V1 tracker with less than 9 points

In the PRO version, it stated that the data output is: "Binocular gaze data, head location in 3D"
- Is this data provided prior to calibration as well? (so that I can create my own calibration function)


Best regards,
Jeffrey

edit:
Version 0.9.56 responses:
{"category":"calibration","request":"start","statuscode":500,"values":{"statusmessage":"Minimum required calibration points is 7"}}
Indicating that The Eye Tribe can be calibrated using 7 points. Tested it and it works.
jeffrey
 
Posts: 1
Joined: 26 Jul 2014, 19:37

Return to General Discussion