Automatic detection of drowsiness in EEG records based on multimodal analysis

Area of Science:

• Neuroscience
• Traffic Safety Engineering
• Signal Processing

Background:

• Driver drowsiness significantly contributes to traffic accidents by impairing attention, danger recognition, and vehicle control.
• Existing methods for drowsiness detection may lack real-time applicability or accuracy.
• Electroencephalography (EEG) signals offer a promising avenue for objective physiological monitoring of alertness levels.

Purpose of the Study:

• To develop and validate an automated method for detecting drowsiness stages using electroencephalography (EEG) data.
• To identify key EEG features that effectively differentiate between alertness and drowsiness.
• To assess the feasibility of implementing this method in real-time vehicle safety systems.

Main Methods:

• EEG data from 18 records were analyzed using time, spectral, and wavelet analysis to extract 19 distinct features.
• A feature selection process, employing the lambda of Wilks criterion, identified 7 optimal parameters.
• A Neural Network classifier was trained and tested using the selected EEG features to distinguish between alertness and drowsiness.

Main Results:

• The automated method achieved high detection rates: 87.4% for alertness and 83.6% for drowsiness.
• The selected 7 EEG-derived parameters demonstrated significant ability to differentiate between alertness and drowsiness states.
• The computed features are computationally efficient, allowing for potential real-time application.

Conclusions:

• The developed method effectively detects driver drowsiness using EEG analysis.
• The identified EEG features provide a reliable basis for an automatic drowsiness detection system.
• This technology holds potential for enhancing vehicle safety and reducing accident rates attributed to driver sleepiness.