Drowsiness Detection with Wireless, User-Generic, Dry Electrode Ear EEG.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Human-Computer Interaction

Background:

• Workplace and driving accidents are often caused by drowsiness.
• Existing drowsiness monitoring systems can be obtrusive or lack accuracy.
• There is a need for discreet and effective drowsiness detection solutions.

Purpose of the Study:

• To develop and evaluate a wireless, multi-channel, dry electrode in-ear electroencephalography (EEG) system for drowsiness monitoring.
• To assess the performance of different machine learning classifiers for drowsiness detection using in-ear EEG data.
• To validate the system's effectiveness across various user-specific and generalized training scenarios.

Main Methods:

• A user study involving five participants and 21 drowsiness trials was conducted.
• A novel in-ear EEG system with dry electrodes and wireless data streaming was utilized.
• Drowsiness detection was implemented using logistic regression, Support Vector Machine (SVM), and random forest classifiers.
• Classifier performance was evaluated using user-specific, leave-one-trial-out, and leave-one-user-out validation strategies.

Main Results:

• The Support Vector Machine (SVM) classifier achieved a 95.9% detection accuracy with user-specific training.
• When tested on unseen users, the SVM model maintained a high accuracy of 94.5%.
• This performance is comparable to state-of-the-art in-ear and scalp EEG drowsiness detection systems.

Conclusions:

• The developed wireless, dry-electrode in-ear EEG system represents a significant advancement in discreet drowsiness monitoring.
• The system demonstrates high accuracy and generalizability, offering a promising solution for reducing accidents.
• This technology has the potential for widespread application in occupational safety and transportation.