Avoiding fatal accidents caused by low vigilance level in driving is very important in our daily lives. Electroencephalography (EEG) has been proved very effective for measuring the level of vigilance. In this paper, we identify light drowsiness state from other states to estimate vigilance level decline by using support vector machine (SVM). Light drowsiness EEG is marked by alpha increasing to 50%. Alert EEG is marked by dominant beta activity and other EEG is labeled as sleep state. Samples of EEG data are trained in SVM program by using 4 features from each frequency band. Mutual information based feature selection method is used to reduce the dimension of features. The accuracy in classification of alert and light drowsiness reaches 91.5% on average.
展开▼
