EEG-based Recognition of Fatigue Driving on Highway

Abstract

To recognize mental fatigue of drivers on highway, a method of EEG signal classification based on wavelet and SVM is presented. EEG signals are decomposed into time-frequency representations using discrete wavelet transform, and as a result the wavelet coefficients in four wavebands namely alpha (a), beta (P), theta (0), delta (8), are obtained. By using the eigenvalues that are composed of energy values and the relative energy values in the four wavebands as training data, two EEG patterns (fatigue driving and non-fatigue driving) are learned by SVM. According to the validating results, the accuracy with which the two states are correctly classified is not sensitive to certain single electrode and is higher in multi-electrode scheme, in which the recognition average accuracy is about 92.24%.