Drowsiness Detection from EEG Spectrum Analysis

Abstract

AbstractDrowsiness is a transient state between consciousness and sleep. In some situations, the operators’ drowsiness raises the risk of major accidents for the lack of agility. Prediction of drowsiness using EEG signals has become the hot topic in the EEG research community. drowsiness states. However, modeling the drowsiness level mathematically from EEG data suffered from the label problems. Drowsiness is subjects’ state of feeling lacking motivation and alertness, which is difficult to be measured at the time of recording the EEG signal. In this paper, we put forward an EEG labeling method employing K-means clustering to separate EEG signal recorded in consciousness and drowsiness states. EEG dataset is divided into two categories according to the EEG rhythmss’ spectrum pattern, and assigned label of drowsiness or consciousness. Comparative study showed thatαand β wave in EEG correlated with the drowsiness level. We also designed a LDA classifier trained with the labeled EEG data, and used it to classify the EEG data into consciousness and drowsiness states. The high classification accuracy illustrates the method put forward in this paper can distinguish these two states (i.e. drowsiness and consciousness) with a high recognition rate.KeywordsDrowsinessEEGK-meansLDA