An Effective Hybrid Deep Learning Model for Single-Channel EEG-Based Subject-Independent Drowsiness Recognition

Area of Science:

• Neuroscience
• Computer Science
• Artificial Intelligence

Background:

• Road accidents are a significant risk, often exacerbated by sleep disorders.
• Existing drowsiness detection methods struggle to fully utilize temporal information from electroencephalogram (EEG) signals.

Purpose of the Study:

• To propose a novel hybrid deep learning model for enhanced drowsiness detection using single-channel EEG signals.
• To improve upon baseline and state-of-the-art deep learning techniques in classifying drowsiness levels.

Main Methods:

• A hybrid model combining discrete wavelet transform (DWT) with long short-term memory (LSTM) and convolutional neural networks (CNN) was developed.
• DWT extracted informative features from EEG sub-bands, while CNN processed spectrogram images to identify drowsiness patterns.
• A majority voting mechanism integrated LSTM and CNN components for robust classification.

Main Results:

• The proposed hybrid model achieved an average accuracy of 74.62% (rounding) and 77.76% (F1-score maximization).
• It demonstrated superior performance compared to conventional methods and other deep learning techniques.
• The model maintained relatively short training and testing times, suitable for real-time applications.

Conclusions:

• The hybrid DWLSTM-CNN model offers a promising solution for accurate and efficient drowsiness detection.
• Feature extraction via DWT and pattern recognition using CNN on spectrograms significantly enhance classification performance.
• This approach holds potential for improving road safety by mitigating accidents related to driver drowsiness.