The influence of eye model parameter variations on simulated eye-tracking outcomes.

Area of Science:

• Ophthalmology
• Biomedical Engineering
• Computer Vision

Background:

• Current eye-tracking research often uses simplified eye models for data simulation.
• These models typically lack consideration for population-based variations in eye biometry.
• This can lead to a disconnect between simulated and real-world eye-tracking performance.

Purpose of the Study:

• To investigate how variations in eye model parameters affect simulated data's predictive accuracy for real-world eye-tracking.
• To determine which biometric variations have the most significant impact on eye-tracking outcomes.

Main Methods:

• Replicated two comparative real-world eye-tracking studies in a simulated environment.
• Employed a high-complexity stochastic eye model with anatomically accurate biometric distributions.
• Analyzed the influence of parameter variations on gaze estimation algorithms.

Main Results:

• Variations in anterior corneal asphericity were found to significantly influence simulated eye-tracking outcomes.
• Both interpolation and model-based gaze estimation algorithms were sensitive to these changes.
• Commonly varied parameters like corneal radius and foveal offset angle showed minimal impact.

Conclusions:

• Realistic simulation of eye-tracking requires incorporating population-specific eye biometry.
• Anterior corneal asphericity is a critical parameter influencing simulated eye-tracking accuracy.
• Future research should focus on more complex, anatomically accurate eye models for robust simulation.