Recently I discovered that the pupil diameter signal can contain artificial "jumps" which are well above the noise level. I guess that these "jumps" are caused by adaptive processes in the camera (automatic exposure control) and/or the eyetracking algorithm (adaptive threshold for pupil edge detection). The amplitude of the jumps seems to be very similar in one recording but can be different across recordings (different subjects, different conditions). The amplitude of these "jumps", I guess, depends on different individual pupil-to-iris contrasts and different levels of focus (also depending on glasses vs. contact lenses vs. no correction, even at the same eye-to-camera distance).
Obviously, having the exposure and threshold controlled adaptively is a good thing, but I wonder if we could have a "tracker" attribute in the API which allows us to freeze/unfreeze the currently used exposure level and threshold in order to get an artifact-free pupil diameter signal when this signal is of particular interest.
.
Pupil diameter of one eye over time. X:500ms/div, Y:2.5pix/div (where "pix" is the unit of "psize" as returned by the TET API).
