What is a blink? Classifying and characterizing blinks in eye openness signals

Area of Science:

• Ophthalmology
• Human-Computer Interaction
• Biomedical Engineering

Background:

• Blinks are crucial for studying human function and behavior.
• Traditional blink detection relies on pupil-size signals, offering indirect eyelid information.
• Existing methods struggle with data quality variations.

Purpose of the Study:

• Compare blink detection using eye openness (EO) signals versus pupil-size (PS) signals.
• Evaluate the robustness of blink detection with varying data quality.
• Introduce a novel, more detailed blink detection method.

Main Methods:

• Developed an algorithm using eye openness (EO) signals to estimate eyelid distance.
• Compared EO blink detection with traditional pupil-size (PS) blink detection algorithms.
• Assessed blink detection accuracy and robustness under high and low data quality conditions.

Main Results:

• High overlap (F1 score: 0.98) in blink rate between EO and PS blinks under optimal conditions.
• Significant overlap (F1 score: 0.94) maintained even with lower data quality.
• EO blinks showed approximately 60 ms longer durations than PS blinks.
• Differences in blink rate were primarily due to pupil-size signal data loss.

Conclusions:

• The eye openness signal provides a more direct and detailed measure of blinks.
• The proposed EO blink detection algorithm enhances accuracy and robustness.
• This method offers advantages for studying blink dynamics and parameters in various research fields.