Gaze3DFix: Detecting 3D fixations with an ellipsoidal bounding volume

Area of Science:

• Cognitive Science
• Computer Vision
• Human-Computer Interaction

Background:

• 2D eyetracking for gaze detection is widespread, but 3D fixation algorithms are lacking.
• Accurate 3D eye movement data is crucial for understanding visual behavior.

Purpose of the Study:

• To present a novel dispersion-based algorithm for estimating 3D fixations.
• To develop and evaluate an open-source toolkit for 3D gaze analysis.

Main Methods:

• A vector-based approach to derive 3D gaze points from 2D data.
• An ellipsoidal bounding volume algorithm to detect 3D fixations.
• Experimental validation using real and virtual stimuli at varying distances.

Main Results:

• The algorithm accurately estimates 3D fixation locations, achieving good congruence with stimulus positions.
• Mean deviation of 3D fixations was 17 mm for both real and virtual stimuli.
• Accuracy showed larger variances at increased stimulus distances (200-600 mm).

Conclusions:

• The Gaze3DFix toolkit offers a ready-to-use solution for 3D eye movement analysis.
• The developed algorithm provides a significant advancement for 3D eyetracking research.
• Open-source availability facilitates broader adoption and further development in the field.