Exploring and Modeling the Effects of Eye-Tracking Accuracy and Precision on Gaze-Based Steering in Virtual...

Area of Science:

• Human-Computer Interaction
• Virtual Reality
• Eye-Tracking Technology

Background:

• Gaze input is increasingly used in Virtual Reality (VR) for intuitive interaction.
• Existing models for steering performance are not directly transferable to gaze-based input due to unique eye movement characteristics and eye-tracking variability.
• A dedicated investigation into gaze-based steering behaviors and performance in VR is needed.

Purpose of the Study:

• To explore and model gaze-based steering behaviors in VR.
• To develop predictive models for user movement time in gaze-based steering tasks.
• To account for the impact of eye-tracking quality on steering performance.

Main Methods:

• Conducted two user studies collecting user behavior data across various path characteristics and eye-tracking conditions.
• Proposed four refined models extending the classic Steering Law for gaze-based steering.
• Incorporated eye-tracking quality as a factor in the predictive models.

Main Results:

• The best-performing model achieved an adjusted R2 of 0.956, improving movement time prediction by 16%.
• Significant reductions in AIC (1132) and BIC (1142) indicate improved model quality and parsimony.
• A second study confirmed the robustness and predictability of the proposed models across different settings.

Conclusions:

• The developed models accurately predict user movement time in gaze-based steering tasks within VR.
• These models offer enhanced prediction capabilities by explicitly considering eye-tracking quality.
• The findings provide a foundation for designing more effective and immersive gaze-based interactions in VR systems.