- 06/02/2015 at 09:56 #2480BarbaraParticipant
I have a question about the meaning of the data returned by the detector which, in my case, is an EyeX. I have performed a very simple test, fixing for 20 s the very center of the screen, keeping the head still, and acquiring the gaze data from the detector. I have then extracted from the data stream provided by the sensor only those data tagged as BOTH_EYES_TRACKED, assuming that the meaning of this tag is that the data provided for left and right eye are valid. I have then evaluated the L-R distance and the corresponding rotation angle (I omit the formulas here, but they are straightforward) and my expectation was to have a reasonably constant value for both the distance and the angle, as well as an angle close to 0° as, during the test, I have tried to keep the eyes as horizontal as possible. What I got instead was that both distance and angle vary over wide intervals of values. I have then repeated the test with other two subjects, getting similar results, even if the spread was less, perhaps because I wear glasses. I would like to know if my expectation is reasonable or not and, in case it is not, what is the exact meaning of the data returned by the sensor, and why they can vary over such a wide interval of values. I have calibrated the detector before each test.
Barbara10/02/2015 at 16:46 #2507AndersParticipant
I’m assuming that you’re using the Gaze SDK with an EyeX Controller, right?
I will try to reproduce your experiment and see if I can come up with an explanation. Be right back.11/02/2015 at 14:54 #2514Patrik [Tobii]Participant
Exactly how do you calculate the distance and rotation angle, and what information on the GazeEyeData struct are you using to calculate this (if you’re using the GazeSDK)?
I wrote a small application that gets the L-R distance and the rotation angle between them, but I cannot see any variations in my data. To be able to reproduce the behavior that you’re getting, all information I can get is valuable.
- You must be logged in to reply to this topic.