Evaluation of Gaze-to-Object Mapping Algorithms for Use in "Real-World" Translatable Neuropsychological Paradigms

Area of Science:

• Virtual Reality (VR)
• Human-Computer Interaction
• Eye-Tracking Technology

Background:

• Eye-tracking is vital for identifying visual attention in virtual reality (VR).
• Applying eye-tracking in VR presents unique challenges for accurate gaze-to-object mapping (GTOM).
• This study evaluates GTOM algorithms in a VR cue-reactivity paradigm.

Purpose of the Study:

• To compare the performance of two gaze-to-object mapping (GTOM) algorithms.
• To assess algorithm accuracy using eye-gaze data from a VR cue-reactivity paradigm.
• To determine optimal object sizes for reliable gaze detection in VR.

Main Methods:

• Participants (n=36) used an HTC Vive Pro Eye in a VR cue-reactivity task.
• Gaze data was analyzed using naive ray-casting versus a combined Tobii G2OM algorithm.
• Gaze accuracy was estimated by percent gaze duration relative to object interaction.

Main Results:

• The combined Tobii G2OM algorithm significantly outperformed naive ray-casting (p=.003).
• Larger objects (>800cm³) showed higher percent gaze duration than smaller objects (≤800cm³; p<.001).
• Both algorithms demonstrated reduced accuracy for objects smaller than 800cm³.

Conclusions:

• Tobii's G2OM algorithm offers superior gaze-to-object mapping in this VR application.
• Object size is a critical factor influencing gaze detection accuracy in VR.
• Recommend utilizing gaze-interactable objects exceeding 800cm³ for enhanced HTC Vive Pro Eye applications.