Recognizing drivers' sleep onset by detecting slow eye movement using a parallel multimodal one-dimensional...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Slow eye movements (SEMs) are reliable physiological indicators of sleep onset in drivers.
• SEMs are often associated with the attenuation of electroencephalogram (EEG) alpha waves.

Purpose of the Study:

• To propose and evaluate a novel Parallel Multimodal 1D Convolutional Neural Network (PM-1D-CNN) model for classifying SEMs.
• To assess the model's effectiveness in detecting sleep onset using electrooculogram (EOG) and EEG signals.

Main Methods:

• A PM-1D-CNN model was developed, employing two parallel 1D-CNN blocks to extract features from EOG and EEG signals.
• Features extracted from both signal types were fused and processed through fully connected layers for classification.
• Model performance was evaluated using both subject-to-subject and cross-subject analyses.

Main Results:

• The PM-1D-CNN model demonstrated superior performance compared to the SGL-1D-CNN and Bimodal-LSTM networks.
• The model achieved high accuracy in classifying SEMs, indicating its robustness.
• Effectiveness was confirmed in both within-subject and across-subject evaluation scenarios.

Conclusions:

• The PM-1D-CNN is an effective deep learning approach for detecting sleep onset via SEM classification.
• This model offers a promising tool for enhancing driver safety systems by identifying drowsiness.
• Multimodal fusion of EOG and EEG data significantly improves sleep onset detection accuracy.