EEG-based assessment of driver's cognitive response in virtual traffic light environment

Traffic accidents linked to the lack of judgment receive an increasing attention due to high fatality rate associated with them. Many of these accidents may be traced to the failure of the drivers to perceive changes in the driving environment, such as changing traffic signals, obstacles, etc. Numerous research initiatives target on driving safety systems in the attempt to improve the road safety. A few emerging reports have proposed the brain computer interface (BCI) approach in designing a driving assistance system that could aid human control by utilizing the driver's cognitive response in the traffic environment.;The purpose of present study was to explore EEG-based signal processing techniques capable of accurate identification of visual evoked potentials elicited by the simulated traffic light signals. Several signal processing techniques were examined to detect the event related potential (ERP) signals obtained for different traffic lights. A bandpass filter was used to remove noise and artifacts; a spatial filter was implemented to minimize the effect of surface currents. Independent component analysis was applied to separate the representative ERP components form the background EEG. Wavelet-based signal decomposition was performed to evaluate the ERPs and compare relative energies of various EEG rhythms between the solid color stimuli and traffic light color stimuli. Analysis of variance (ANOVA) was performed to assess the energy densities of different EEG rhythms. The cleaned and denoised EEG signals were classified using three algorithms---namely, Euclidean distance, Correlation, and Neural Network to identify the EEG pattern associated with each traffic light. The best classification accuracy of 57 % was achieved by using neural network. The work demonstrates the feasibility of detecting and analyzing the ERP signals that correspond to the driver's cognitive responses to the traffic light colors.