Eye tracker calibration: How well can humans refixate a target?

Area of Science:

• Ophthalmology
• Computer Science
• Human-Computer Interaction

Background:

• Traditional pupil-based eye trackers have inherent inaccuracies, often around 0.5 degrees.
• Factors influencing gaze signal accuracy include pupil-size changes and human refixation capabilities.
• The pupil-size artefact (PSA) is a known confounder in pupil-based eye tracking.

Purpose of the Study:

• To investigate the impact of pupil-size changes (PSA) on eye-tracking accuracy.
• To evaluate the role of human refixation precision in limiting gaze signal accuracy.
• To assess the effectiveness of PSA correction methods for improving eye-tracking quality.

Main Methods:

• Experiment 1: Utilized a traditional pupil-based eye tracker to measure inaccuracies.
• Experiment 2: Employed a retinal eye tracker (FreezeEye) to eliminate pupil-based estimation.
• Assessed gaze accuracy during calibration and validation phases in both experiments.

Main Results:

• The pupil-size artefact (PSA) was found to significantly affect gaze accuracy, causing up to 0.5 degrees of error.
• Implementing corrections based on pupil size and gaze shift relationships substantially reduced these inaccuracies.
• Experiment 2 confirmed that humans possess precise refixation abilities, not limiting gaze signal accuracy.

Conclusions:

• Pupil-size changes and the associated artefact are major contributors to eye-tracking inaccuracy.
• Corrective strategies targeting the PSA can significantly enhance the quality of eye-tracking data.
• Human refixation precision is not a limiting factor for accurate gaze signal detection.