Deep Learning for Detecting Multi-Level Driver Fatigue Using Physiological Signals: A Comprehensive Approach.

Area of Science:

• Engineering
• Computer Science
• Neuroscience

Background:

• Driver fatigue is a significant cause of traffic accidents.
• Existing methods for fatigue detection often rely on self-reporting, which can be unreliable.

Purpose of the Study:

• To develop and evaluate a new approach for multi-level driver fatigue detection.
• To create a robust system for identifying and classifying different states of driver tiredness.

Main Methods:

• A multi-level driver tiredness database was created using physiological signals (ECG, EEG, EMG, respiratory effort).
• A customized end-to-end architecture combining adversarial generative networks and convolutional neural networks was employed for feature extraction and classification.
• Type 2 fuzzy sets were utilized within the architecture to mitigate uncertainty, replacing traditional activation functions.

Main Results:

• The system achieved high accuracy in detecting driver fatigue across different levels: 96.8% for two-level, 95.1% for three-level, and 89.1% for five-level fatigue detection.
• The use of EEG signals for processing, corroborated by other physiological data, provided a reliable fatigue assessment.
• The investigation into type 2 fuzzy sets demonstrated their effectiveness in enhancing model performance.

Conclusions:

• The proposed model demonstrates optimal performance in identifying various levels of driver fatigue with high accuracy.
• This advanced driver fatigue detection system has significant potential for practical implementation in vehicle safety systems to warn drivers.
• The integration of physiological signals and sophisticated AI offers a promising direction for future driver monitoring technologies.