Classifying Driving Fatigue Based on Combined Entropy Measure Using EEG Signals

Driving fatigue is a common occupational hazard for any long distance or professional driver, and fatigue detecting has major implications for transportation safety. Monitoring physiological signal while driving can provide the possibility to detect the fatigue and give the necessary warning. In this paper, fifty subjects participated in driving simulations experiment with their recorded EEG signals to induce two kinds of fatigue states: Alert and drowsy. Two nonlinear methods, approximate Entropy (AE) and Sample Entropy (SE), were used to characterize irregularity and complexity of EEG data. Subsequently Support Vector Machine (SVM) was applied to classify these two fatigue states. The experimental result shows that two complexity parameters are significantly decreased as the fatigue level increases. The result indicates that both of two nonlinear indicators can be used to characterize driver fatigue level. Furthermore, the combined measure feature results in higher classification accuracy, indicating the proposed classification method is more robust and effective, compared with single complexity measure.