Calibration-Free Mobile Eye-Tracking Using Corneal Imaging

Area of Science:

• Computer Vision
• Human-Computer Interaction
• Biomedical Engineering

Background:

• Traditional eye-tracking systems often require cumbersome calibration procedures.
• Mobile eye-tracking demands unobtrusive and continuous gaze estimation.
• Accurate gaze tracking is crucial for various applications, including augmented reality and user behavior analysis.

Purpose of the Study:

• To present and evaluate a novel mobile eye-tracking system that eliminates the need for manual calibration.
• To enable continuous and reliable gaze estimation in real-world, dynamic environments.
• To achieve high accuracy comparable to state-of-the-art methods without user intervention.

Main Methods:

• A mobile device integrating three cameras: IR eye camera, RGB eye camera, and front-scene RGB camera.
• Development of a corneal imaging system for robust gaze estimation.
• Implementation of deep learning algorithms and 3D geometric computations for automatic per-user system calibration.
• Automatic computation of point-to-point transformation by matching corneal and scene images.

Main Results:

• The system successfully auto-calibrates, allowing users to start tracking immediately.
• Demonstrated reliable gaze point estimation in both indoor and outdoor real-life scenarios.
• Achieved an average gaze error of 1.6° indoors and 1.69° outdoors.
• Performance is competitive with existing state-of-the-art eye-tracking solutions.

Conclusions:

• The developed calibration-free mobile eye-tracking system offers a significant advancement in usability and accessibility.
• The system provides accurate and reliable gaze estimation suitable for diverse real-world applications.
• This approach paves the way for more natural and intuitive human-computer interaction through seamless eye-tracking technology.