A subject-transfer framework for obviating inter- and intra-subject variability in EEG-based drowsiness detection.

Area of Science:

• Neuroscience
• Computer Science
• Biomedical Engineering

Background:

• Brain-computer interfaces (BCIs) using electroencephalogram (EEG) face challenges due to inter- and intra-subject variability.
• Current BCI training is time-consuming, limiting real-world applications like drowsiness detection.

Purpose of the Study:

• To assess variability in EEG data for BCI applications.
• To develop and validate a subject-transfer framework for EEG-based drowsiness detection.
• To reduce the calibration time for new BCI users.

Main Methods:

• Applied hierarchical clustering to analyze inter- and intra-subject variability in a large EEG dataset from a simulated driving task.
• Developed a subject-transfer framework utilizing a model pool and minimal alert baseline data for new users.
• Compared the proposed framework with the conventional within-subject approach.

Main Results:

• The proposed framework reduced calibration time by 90% (from 18.00 min to 1.72 ± 0.36 min) compared to conventional methods.
• Performance was maintained without significant compromise (p=0.0910).
• Demonstrated the feasibility of transferring EEG-based drowsiness detection models across subjects.

Conclusions:

• The developed subject-transfer framework offers a practical solution for plug-and-play drowsiness detection.
• This approach significantly minimizes user-specific calibration, paving the way for broader BCI applications.
• The findings support the advancement of real-world BCI systems by addressing data variability challenges.