Trajectory prediction of saccadic eye movements using a compressed exponential model.

Area of Science:

• Vision Science
• Neuroscience
• Computer Science

Background:

• Gaze-contingent display paradigms are crucial in vision research.
• Time delays in eye-tracking systems can cause misalignment between gaze and displayed information, compromising experimental accuracy.

Purpose of the Study:

• To develop and evaluate a method for reducing gaze-contingent display misalignment.
• To improve the accuracy of eye-tracking systems by predicting saccadic eye movements.

Main Methods:

• Utilized a compressed exponential function to model saccadic eye movement trajectories.
• Evaluated the algorithm using experimental data from 1,212 saccades recorded with EyeLink 1000 and Dual-Purkinje Image (DPI) eye trackers.
• Assessed the impact of a 10-millisecond time delay on saccade trajectory prediction and misalignment reduction.

Main Results:

• The model demonstrated high agreement (R² > 0.96) in fitting eye displacement.
• Saccade trajectory prediction reduced misalignment by 30-60% (EyeLink) and 20-40% (DPI) for saccades > 8° with a 10ms delay.
• In a simulated 100-Hz display, prediction reduced misalignment > 2° from 45% to 20% (EyeLink) and 42% to 26% (DPI).

Conclusions:

• The proposed saccade-prediction algorithm effectively reduces gaze misalignment in eye-tracking systems.
• This method enhances the accuracy of gaze-contingent displays, particularly for larger saccades.
• Further research may explore optimizations for small saccades and early saccade stages.