Snip-and-Stitch: A Simple, Universal Correction for the Pupil Foreshortening Error

Abstract

When using video-based eye tracking devices to measure pupil size, a measure of cognitive load and arousal, gaze-angle relative to the tracker influences reported pupil size. This happens due to the view of the pupil becoming more obscured as the eye rotates away from the camera. Existing correction methods rely on extensive, complicated calibration for each participant, leaving room for measurement error. Additionally, these methods only work for a limited set of eye tracking devices. In this paper we introduce the ‘Snip-and-Stitch’ method, which corrects the PFE by ‘snipping’ pupil size change during a saccade, and ‘stitching’ pupil size back by subtracting the difference between pre- and post-saccadic pupil size. This novel method reduced the pupil foreshortening error after one saccade by 85% but was found to accumulate a positive error which surpasses the average error by the PFE after 8 saccades on average. Research where participants make only a few saccades per trial are suggested as research where applying the snip-and-stitch method has clear benefits over existing methods. Improvement upon the snip-and-stitch method is still possible, for instance in intra-saccadic pupil size change estimation, as well as saccade endpoint detection.