Gaze Error Estimation and Linear Transformation to Improve Accuracy of Video-Based Eye Trackers

Area of Science:

• Computer Vision
• Human-Computer Interaction
• Psychology

Background:

• Eye tracking technology is essential across diverse disciplines including psychology, medical training, marketing, and human-computer interaction.
• A persistent challenge in eye tracking is maintaining high accuracy, especially over large fields of view, under both free-viewing and head-stabilized conditions.

Purpose of the Study:

• To introduce and evaluate a straightforward method for enhancing the accuracy of video-based eye trackers.
• To address the limitations of current eye tracking systems in achieving precise gaze detection over extensive viewing areas.

Main Methods:

• The proposed approach involves applying linear coordinate transformations, such as stretching, shearing, and translation, to correct gaze accuracy errors.
• These transformations are used to re-calibrate video-based eye trackers, aiming to improve their performance.

Main Results:

• Re-calibration using linear transformations demonstrably improves the accuracy of video-based eye trackers.
• The effectiveness of this method was validated across a large field of view, indicating a significant enhancement in gaze accuracy.

Conclusions:

• Linear coordinate transformations offer a simple yet effective solution for improving the accuracy of video-based eye trackers.
• This technique holds promise for advancing applications that rely on precise eye tracking over wide fields of view.