EEG-Based Multiple Entropy Analysis for Assessing Driver Fatigue

Abstract

For new automatic technology, an EEG-based approach for studying driver fatigue is one of the potential important research field in traffic safety. In this article, the proposed method based on EEG signals aimed to assess driver fatigue by using multi-entropy measures and compare the performance with channel combination and multiple classifiers. Given that EEG signals are unstable and non-linear, that using several common entropy evaluators to analyze EEG is more appropriate, including spectral entropy, approximate entropy, sample entropy and fuzzy entropy. In this paper, unlike other methods using whole electrodes and single classifier, the influence of channel combination on fatigue detection is discussed, and three types of common classifiers including Random Forest, Decision Tree and K-Nearest Neighbor are applied for classifying driver fatigue, implying that a comprehensive comparison is deeply discussed among them. A simulated driving experiment in this study for twenty-two healthy adults was used to perform continuous signal acquisition for about 20 minutes. The experimental results show that the proposed method can hit the highest accuracy for driver fatigue detection of 97.5% with the leave-one-out cross-validation approach, implying that it could be suitable for accessing driver fatigue by using four entropy measures based on O1 channel and RF classifier.