PNN based driver drowsiness level classification using EEG

Abstract

In this work, we classify the driver drowsiness level (awake, drowsy, high drowsy and sleep stage1) based on different wavelets and probabilistic neural network classifier using wireless EEG signals. Deriving the amplitude spectrum of four different frequency bands delta, theta, alpha, and beta of EEG signals. Comparing the results of PNN based on spectral centroid, and power spectral density features extracted by different wavelets (db4, db8, sym8, and coif5) from the amplitude spectrum of the said bands. As results of this study indicates that the best average accuracy achieved of 61.16% based on power spectral density feature extracted by db4 wavelet.