Real driving environment EEG-based detection of driving fatigue using the wavelet scattering network

Area of Science:

• Neuroscience
• Machine Learning
• Traffic Safety

Background:

• Driving fatigue is a primary cause of traffic accidents.
• Accurate and rapid detection of driver fatigue is essential for road safety.

Purpose of the Study:

• To develop a precise method for detecting driving fatigue in real-world driving conditions.
• To evaluate the effectiveness of the Wavelet Scattering Network (WSN) algorithm for fatigue classification.

Main Methods:

• Collected electroencephalogram (EEG) signals from 12 subjects in a real driving environment.
• Classified EEG signals into 'fatigue' and 'awake' states.
• Utilized the WSN algorithm to extract wavelet scattering coefficients from EEG signals.
• Employed a Support Vector Machine (SVM) with WSN coefficients as feature vectors for classification.

Main Results:

• Achieved an average classification accuracy of 99.33% for the 12 subjects.
• Demonstrated high average precision (99.28%), recall (98.27%), and F1 score (98.74%).
• Validated performance on the SEED-VIG dataset, yielding 99.39% accuracy and high precision, recall, and F1 scores.

Conclusions:

• The WSN algorithm shows superior classification accuracy compared to CNN, SDBN, STCNN, and LSTM.
• The WSN method offers good versatility, effective recognition with small datasets, and fast processing for real-time monitoring.
• WSN is a promising approach for real-time driver fatigue detection, enhancing traffic safety.