Automatic processing of gaze movements to quantify gaze scanning behaviors in a driving simulator

Area of Science:

• Human-Computer Interaction
• Neuroscience
• Transportation Engineering

Background:

• Driving requires continuous environmental scanning using eye and head movements.
• Approaching intersections involves significant lateral gaze scans for situational awareness.
• Manual analysis of gaze data in driving studies is time-consuming and labor-intensive.

Purpose of the Study:

• To introduce and evaluate the Gaze Scan Algorithm (GSA) for automated quantification of lateral gaze scans during driving.
• To assess the accuracy and efficiency of the GSA compared to manual annotation.
• To provide a tool for enhanced understanding of visual scanning behavior in driving.

Main Methods:

• Development of the Gaze Scan Algorithm (GSA) to detect lateral saccades and merge them into gaze scans.
• Algorithm identifies scan start/end points in time and eccentricity.
• Validation using gaze data from a high-fidelity driving simulator, comparing GSA output to expert-annotated ground truth.

Main Results:

• The Gaze Scan Algorithm (GSA) successfully identified 96% of ground truth gaze scans.
• The algorithm demonstrated high accuracy in quantifying the magnitude and duration of gaze scans.
• Performance metrics were comparable to inter-expert coder variability, indicating reliability.

Conclusions:

• The Gaze Scan Algorithm (GSA) offers a reliable and efficient automated method for analyzing gaze scan data.
• It significantly reduces the time required for manual annotation in driving simulation studies.
• The GSA enhances eye tracking and mobility research by providing detailed quantification of visual scanning patterns.