Vigilance analysis based on EEG band power using Support Vector Machine

Abstract

Vigilance analysis associated with safe driving based on EEG has drawn considerable attention of researchers in recent years. Preventing traffic accidents caused by low level vigilance is highly desirable. This paper presents a novel vigilance analysis system by evaluating electroencephalographic (EEG) changes. EEG signals are preprocessed with independent component analysis to eliminate noise from the original EEG recording. Then, EEG band power features are extracted by using Fast Fourier Transform (FFT). These features serve as an input for further classification. Support Vector Machine (SVM) is subsequently employed as a classifier to distinguish vigilance level. Nine healthy subjects participated in our experiment at which they drive a car in driving simulator. Experimental results reveal that the proposed approach could be used to develop a noninvasive monitoring system for vigilance state.