Measuring vigilance decrement using computer vision assisted eye tracking in dynamic naturalistic environments

Area of Science:

• Computer Vision
• Human-Computer Interaction
• Cognitive Psychology

Background:

• Eye tracking is valuable for cognitive performance monitoring in real-world settings.
• Dynamic environments pose challenges for accurate eye tracking due to stimulus variability.
• Investigating novel computer vision approaches is crucial for advancing eye tracking in complex scenarios.

Purpose of the Study:

• To develop a computer vision-assisted eye tracking analysis method using fixations.
• To define and validate a 'fixation score' for quantifying visual attention dynamics.
• To assess the relationship between the fixation score and cognitive performance metrics like reaction time.

Main Methods:

• Collected eye movement data from a long-duration naturalistic driving experiment.
• Utilized the Source Invariant Feature Transform (SIFT) algorithm to identify Areas of Interest (AOIs).
• Defined a 'fixation score' to measure the distribution of gaze between target and non-target AOIs.

Main Results:

• A statistically significant negative correlation (r = -0.2253, p < 0.05) was observed between the fixation score and reaction time.
• The fixation score decreases as vigilance declines and visual attention shifts away from target objects.
• Higher fixation scores indicate greater focus on target AOIs, correlating with shorter reaction times.

Conclusions:

• The proposed 'fixation score' effectively reflects changes in visual attention during driving.
• This computer vision-assisted eye tracking method shows promise for monitoring cognitive performance and vigilance.
• The findings have implications for driver safety and the development of advanced driver-assistance systems.