Estimating User Gaze Depth Perception in Real -Time for Extended Reality Environments

Abstract

The progression of head-mounted displays (HMDs) to incorporate eye tracking in extended reality settings enables opportunities to develop adaptive HMDs capable of responding to user context and state to provide relevant information faster, which can result in faster training and better decision-making. The purpose of this research is to develop a method that allows for real-time estimation of perceptual depth in extended reality settings using eye-tracking data. In three different experimental conditions (virtual reality, augmented reality,and real), 13 human subjects fixated on targets at four varying depths while eye-tracking data was collected. Here, fixation periods are isolated and segmented in the eye-tracking data and the average inverse eye vergence angle (EVA) and interpupillary distance (IPD) are calculated. Using these two variables as features and the depth as the target, support vector machine (SVM), random forest, and XGBoost models exhibited mean classification accuracy of 50.1 , 48.9 , and 49.0 , respectively. Near/far classification for distances of 0.25 m versus 4.0 m yielded classification accuracies of 61.0 , 85.4 , and 85.6 for SVM, random forest, and XGBoost, respectively. We conclude that user perceptual depth can be estimated using EVA and IPD and adaptive HMDs are feasible using similar models for gaze depth estimation.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Oct 30, 2023
Accession Number
AD1214025

Entities

People

  • Nathan Villavicencio
  • Russell Cohen Hoffing

Organizations

  • California State University
  • United States Army Research Laboratory

Tags

Readers

  • Adaptive Control and Estimation with Uncertainty in Dynamic Systems.
  • Human-Computer Interaction (HCI).
  • Vision Science/Vision Psychology/Cognitive Neuroscience.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference