Enhancing accuracy of mental fatigue classification using advanced computational intelligence in an...

Area of Science:

• Neuroscience
• Transportation Safety
• Biomedical Engineering

Background:

• Mental fatigue in drivers poses a significant safety risk.
• Current methods for detecting driver fatigue have limitations.
• Electroencephalography (EEG) offers a potential avenue for objective fatigue monitoring.

Purpose of the Study:

• To develop and evaluate a system for classifying mental fatigue and alert states in drivers using EEG signals.
• To compare the effectiveness of different EEG channel configurations for fatigue detection.
• To assess the feasibility of using fewer EEG channels for practical fatigue monitoring.

Main Methods:

• Feature extraction using autoregressive (AR) model-based power spectral density (PSD).
• Classification employing a fuzzy particle swarm optimization with cross-mutated artificial neural network (FPSOCM-ANN).
• Utilized 32-channel EEG data and evaluated a reduced set of eleven frontal EEG channels.

Main Results:

• The 32-channel EEG system achieved an overall accuracy of 80.51%, with specificity at 82.02% and sensitivity at 78.99%.
• A reduced configuration using eleven frontal EEG channels demonstrated comparable performance, achieving 75.65% accuracy.
• The eleven-channel system showed improved specificity (77.52%) and sensitivity (73.78%) compared to previous studies.

Conclusions:

• The proposed FPSOCM-ANN system effectively classifies mental fatigue states in drivers using EEG signals.
• A reduced number of frontal EEG channels provides a practical and ergonomic solution for fatigue monitoring.
• This approach holds promise for enhancing driver safety through real-time fatigue detection.