Eye-gaze calibration autocorrection using interface design elements

Improving calibration of infrared and webcam-based gaze trackers by leveraging natural gaze patterns from preattentive visual features.

Initial situation

Eye-gaze trackers are widely used for research, accessibility, and interaction, but require frequent calibration. This process interrupts workflow and reduces usability. Preattentive visual features, such as high-contrast colors, motion, grouping, and large element size, naturally attract human attention and produce predictable gaze patterns, which could be used for implicit calibration.

Problem statement / Project goal

This project investigates whether preattentive visual features can be leveraged to implicitly calibrate gaze trackers during normal use. The aim was to determine which features reliably attract gaze, how they can be integrated into real-world websites, and how environmental factors influence calibration accuracy.

Solution developed and its benefits

The solution consisted of two experiments and a correction algorithm test:

• Identified the most effective preattentive visual features for attracting gaze without disrupting task flow.
• Demonstrated that these features can serve as implicit calibration points on real websites.
• Developed a correction algorithm to improve gaze accuracy using identified gaze patterns.
• Investigated the influence of lighting, background, and display setup on calibration effectiveness.

Results show that implicit calibration using preattentive visual features is a viable way to enhance gaze tracker usability and reduce interruptions for the user.

Key terms

• Preattentive Visual FeaturesDesign elements that naturally attract gaze.
• Implicit CalibrationImproving tracker accuracy without explicit recalibration.
• WebGazerJavaScript-based webcam eye-tracking library.
• Infrared TrackingHigh-precision gaze tracking for baseline comparison.

Team