Classification of Driver Daydreaming Using Data Mining Techniques

Abstract

Driver distraction is one of the leading causes of car accidents. Daydreaming, as a common form of cognitive distraction, is difficult to detect and often overlooked. This research studies the application of data mining techniques in classification of drivers daydreaming and non-daydreaming through a quantitative analysis of electroencephalogram (EEG) recordings. EEG Data were gathered from simulated driving tasks. Features were extracted from EEG frequency power bands on six electrode sites. Two data mining techniques, namely, Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM), were employed to classify EEG samples associated with daydreaming and non-daydreaming situation. The best testing performance for the in-subject classification had a sensitivity of 60.04% and a specificity of 64.59% by SVM. The receiver operating characteristic (ROC) analysis revealed that the averaged value of the area under ROC curves of LDA and SVM for most of the subjects were both greater than 0.60. The classification results of this study indicated that it is possible to detect driver daydreaming based on EEG signal. Potential applications of this paper include the design of adaptive in-vehicle system and evaluation of driver daydreaming.