Home Forums Eye Tracking Devices About the sampling rate of tobii Reply To: About the sampling rate of tobii

#21514
Grant [Tobii]
Keymaster

Hi @elta, thank you for your query. The issue you raise is less to do with sampling frequency but rather geometry.

When the eye is not looking straight towards the camera, the pupil appears as an ellipse, and there is refraction in the cornea as well that complicates things.
Tobii does a better job than most trackers since we do not estimate pupil directly from the image, but we reconstruct a complete 3d model with cornea surface, refraction and all. From this model, we extract the actual pupil diameter in millimeters (as opposed to the apparent pupil diameter in pixels that most other trackers do).

It is not perfect, for several reasons. Imperfections in the image is not the major problem. imperfection in the geometric model is a bigger source of errors.
In the 4C for example, some images are BP and some images are DP. This is also a source of problems. the contrast of the pupil edge is different for bright and dark pupil images. Stereo camera Pro trackers (Fusion and Spectrum) does not cycle between DP and BP like the consumer trackers partly to avoid this problem.

Let us know if you have any further queries. Best Wishes.