ANN based Joint Time and frequency analysis of EEG for detection of driver drowsiness

Abstract

Drowsiness detection plays a vital role in accidents avoidance systems, thereby saving many precious lives. According to the World Health Organization, Drowsiness has been the radical contributor of road fatalities. Electroencephalogram (EEG) is a physiological signal which relays the functioning of Brain and widely used in the diagnosis of Neurological Disorders. This study extrapolates the EEG signal analysis to examine several cognitive tasks. In this report, the EEG signal is processed to detect the behavioural patterns of the brain and drowsiness state of the drivers while performing monotonous driving for long distances. An eight-channel EEG data acquisition system is used to acquire the EEG data from 20 male volunteers. The EEG signal is pre-processed and decomposed into various rhythms by applying Digital filter in MATLAB 2007b (Mathworks, Inc., USA). Time-Frequency Domain analysis has been done to extract certain features PSG and PRMSD which are statistically significant (ρ < 0.05) in the detection of drowsiness. The driving profile is classified into Active and Drowsy by a threshold, and linear regression analysis has been performed on the features extracted. A Drowsiness index is proposed stating a positive correlation (0.8-0.9) between the Total mean and the drowsy mean of the subject.