Portable Drowsiness Detection through Use of a Prefrontal Single-Channel Electroencephalogram

Area of Science:

• Neuroscience
• Biomedical Engineering
• Human-Computer Interaction

Background:

• Drowsiness detection is crucial for driver safety, product evaluation, and workplace monitoring.
• Conventional electroencephalogram (EEG) devices are costly and restrictive, limiting practical applications.
• There is a need for accessible, portable EEG systems to assess drowsiness in real-world scenarios.

Purpose of the Study:

• To develop and evaluate a low-priced, single-channel, prefrontal electroencephalogram (EEG) system for drowsiness detection.
• To assess the feasibility of using portable EEG devices for practical drowsiness estimation.
• To compare feature extraction methods for EEG-based drowsiness detection.

Main Methods:

• Developed a drowsiness detection system using a low-cost, prefrontal single-channel EEG device.
• Compared power spectral density (PSD), autoregressive (AR) modeling, and multiscale entropy (MSE) for EEG feature extraction.
• Utilized stepwise linear discriminant analysis (SWLDA) for PSD selection and support vector machine (SVM) with radial basis function (RBF) kernel for pattern recognition.
• Optimized SVM hyperparameters using grid search and evaluated performance with 10-fold cross-validation.

Main Results:

• The developed system achieved 72.7% classification accuracy for drowsiness detection using PSD with SWLDA and SVM.
• A practical experiment demonstrated a significant difference between drowsiness-evoked and concentration-needed tasks (*p < 0.05).
• The portable EEG system effectively quantified drowsy states in a practical setting.

Conclusions:

• The low-priced, portable EEG drowsiness detection system is effective for quantifying drowsy states.
• This system holds potential for practical applications in education, entertainment, and workplace environments.
• Portable EEG technology can be successfully applied in research and business settings for drowsiness monitoring.