Combined eye activity measures accurately estimate changes in sustained visual task performance

Area of Science:

• Human-Computer Interaction
• Cognitive Psychology
• Biomedical Engineering

Background:

• Fatigue significantly impacts performance in sustained tasks, necessitating reliable monitoring methods.
• Traditional performance metrics may not capture rapid, fatigue-related changes in real-time.
• Eye activity offers a potential window into cognitive and physical states affecting performance.

Purpose of the Study:

• To model fatigue-related performance decrements during a visual compensatory tracking task.
• To investigate the efficacy of various eye activity measures in predicting tracking performance.
• To develop individualized, real-time performance estimation models.

Main Methods:

• Utilized video-based eye tracking to capture blink duration/frequency, fixation dwell time/frequency, and pupil diameter.
• Employed non-linear regression and artificial neural network techniques for data analysis.
• Cross-validated individualized models on separate testing sessions.

Main Results:

• Mean tracking error increased monotonically in the initial 11 minutes, followed by a plateau.
• A general regression model using fixation data achieved R=0.68.
• Individualized neural network models yielded the highest correlation (R=0.82) and lowest RMS error (mean=1.23 disk radii).

Conclusions:

• Multiple eye activity measures, particularly fixation patterns, can be combined to estimate tracking performance.
• Individualized neural network models offer superior accuracy for real-time performance prediction.
• Eye-based measures show promise for monitoring performance during prolonged, demanding tasks.