Early driver fatigue detection from electroencephalography signals using artificial neural networks

Area of Science:

• Neuroscience
• Artificial Intelligence
• Transportation Safety

Background:

• Driver fatigue is a significant safety concern in transportation.
• Existing driver fatigue detection methods often lack accuracy and efficiency.
• Electroencephalogram (EEG) signals offer a promising physiological measure for fatigue assessment.

Purpose of the Study:

• To develop and evaluate an artificial neural network (ANN) based system for driver fatigue detection.
• To investigate the efficacy of a novel training optimization technique, the magnified gradient function (MGF), for ANN-based fatigue detection.
• To compare the performance of the proposed system in classifying fatigue for both professional and non-professional drivers.

Main Methods:

• Acquisition of electroencephalogram (EEG) data from 20 professional truck drivers and 35 non-professional drivers.
• Processing of time-domain EEG data into alpha, beta, delta, and theta frequency bands.
• Implementation of an artificial neural network (ANN) utilizing the magnified gradient function (MGF) for training optimization.
• Modification of the standard back propagation (SBP) algorithm to create the MGF technique for reduced training time.

Main Results:

• The ANN system achieved 81.49% accuracy in classifying professional driver fatigue (80.53% sensitivity, 82.44% specificity).
• The system demonstrated 83.06% accuracy for non-professional driver fatigue (84.04% sensitivity, 82.08% specificity).
• The magnified gradient function (MGF) significantly reduced the training time required for the neural network.

Conclusions:

• The developed ANN system effectively detects driver fatigue using processed EEG data.
• The MGF training optimization technique enhances the efficiency of ANN models for fatigue detection.
• The system shows robust performance in distinguishing fatigued states across different driver groups.