Improving Performance of the Human Pupil Orbit Model (HPOM) Estimation Method for Eye-Gaze Tracking

Area of Science:

• Biomedical Engineering
• Computer Vision
• Human-Computer Interaction

Background:

• Eye-gaze direction-tracking technology is crucial in diverse fields like medicine, education, engineering, and gaming.
• Current systems require enhanced response speed, stability, accuracy, and precision.
• Existing human pupil orbit model (HPOM) estimation methods face limitations in speed and system load.

Purpose of the Study:

• To propose a novel method for improving the speed of eye-gaze direction tracking.
• To reduce system load and maintain high precision in HPOM estimation.
• To enhance the overall performance of eye-gaze tracking systems.

Main Methods:

• Developed a new HPOM estimation method based on the consistent orientation of the minor axis of elliptical-deformed pupils towards the rotational center.
• Utilized simulation experiments to validate the proposed method's performance.
• Analyzed speed improvements, accuracy (pixel error on x and y axes), and precision (0.0 pixels for ES ≤ 7).

Main Results:

• Achieved a speed improvement of at least 74 times, completing estimation in under 7 ms.
• Demonstrated high accuracy with maximum errors of approximately 0.2 pixels (x-axis) and 8 pixels (y-axis).
• Showcased precision of 0.0 pixels for 7 or fewer estimation samples (ES), comparable to existing HPOM methods.

Conclusions:

• The proposed method offers a significant speed enhancement for HPOM estimation with minimal impact on accuracy and precision.
• The method's conservative performance against allowable angle error (AAE) suggests robustness under worst-case conditions.
• HPOM can be accurately and precisely estimated by adjusting AAE based on system performance and environmental factors.