Driver State Analysis for ADAS using EEG Signals

Abstract

This paper discusses about the driver state analysis for Advanced Driver Assistance System (ADAS) application in automotive and check whether the driver is drowsy or alert. Driver state analysis is one of the important tasks in today's automotive. In this paper, EEG signals are used for driver state analysis. The sleep data set for analysis is obtained from Physionet. The study of characteristics of EEG signals and its different frequency rhythms like gamma, beta, alpha, theta and delta. The drowsy state indication is observed in alpha and theta frequency rhythms. The data is pre-processed to remove artifacts and features representing drowsiness are extracted. The feature reduction techniques are used to reduce the humongous features whose computation time is more during classification. ADAS application should produce quick response for the analysis of EEG signal, hence the algorithm used to classify the driver state (drowsy/alert) should have less computational time. The comparison of computational time and accuracy of different classification algorithms like SVM and logistic regression is done and the best algorithm with less computational time and more accuracy is selected. The proposed method is used to analyse the driver state and further to analyze different sleep stages for clinical purposes.