We proposed an electroencephalographic (EEG) signal analysis approach to investigate the driver's cognitive response to traffic-light experiments in a virtual-reality-(VR-) based simulated driving environment. EEG signals are digitally sampled and thentransformed by three different feature extraction methods including nonparametric weighted feature extraction (WEE),principal component analysis (PCA), and linear discriminant analysis (LDA), which were also used to reduce the feature dimensionand project the measured EEG signals to a feature space spanned by their eigenvectors. After that, the mapped data could beclassified with fewer features and their classification results were compared by utilizing two different classifiers including k nearestneighbor classification (KNNC) and naive bayes classifier (NBC). Experimental data were collected from 6 subjects and theresults show that NWFE+NBC gives the best classification accuracy ranging from 71%-77%, which is over 10%-24% higherthan LDA+KNN1. It also demonstrates the feasibility of detecting and analyzing single-trial EEG signals that represent operators'cognitive states and responses to task events.
展开▼
