- This topic has 3 replies, 2 voices, and was last updated 6 years, 2 months ago by
.
- Posts
hi,
i calibrated my EyeTracker 4C via the tray app, and wanted to test it’s results.
i drew a point on screen center (940, 520) and focused on it for few seconds.
i immediately closed my eyes, and then calculated the average of the points, and drew their point.
the result, time after time was shocking, i got deviation of about 5-10px on X and 35-50px on Y – and the point center i wanted to recognize as (940, 520) was actually recognized as (930, 470) and was drawn there as well.this happens time after time and for different people and different screens…
i use the tobii_stream_engine nugget & the results were pretty similar on both gaze_point stream and gaze_data stream (i averaged both eyes gazing points)
what can be the problem and how can i overcome it?
Hi @tzaccai, gaze data from any eye tracker is inherently noisy due mostly to the natural small movements of the eye even when fixating on a specific point. Indeed, it is for this reason that eye tracking applications depend on fixations rather than individual gaze points to determine the most likely locations a user is looking.
At a distance of around 30cm between the tracker and the user, one might anticipate a deviation of pixels up to around 10/12 pixels so your case is somewhat outwith nominal accuracy in the y dimension.
Can you kindly try the ‘gaze trace’ feature available within the Tobii Core Software Tray icon and let me know what kind of accuracy you experience when looking around the screen.. the size of the bubble indicates the approximate accuracy within which
the gaze is deemed to be looking. In addition, you can also try ensuring that the eye tracker is not close to any major sources of infrared light such as large open windows or spotlights in the periphery.hi @Grant & thanks a lot!
i’ve turned the gaze trace on, and saw that my dot (or words when i try to read) are within the lower part of the bubble. and that is after calibration.
i must say i don’t sit within 30cm from the tracker & screen, but more close to 80cm, and there is a window behind me – but it is dark now 🙂i’m using the gazing point stream, but try to calc fixations by myself because the fixation stream of the SDK is not as accurate as well… and i think i’m doing nice with that 😉
Hi @tzaccai, it would be very interesting to see what methodology you are using to develop your own fixation algorithms and ideally share the code if possible? For your interest, you may find other examples of custom fixation filters to try @ https://github.com/davebraze/FDBeye/wiki/Researcher-Contributed-Eye-Tracking-Tools
Hopefully that information is of some use to you!
- You must be logged in to reply to this topic.