EEG based detection of alcoholics using spectral entropy with neural network classifiers

This paper suggests the application of gamma band spectral entropy for the detection of alcoholics. First, the gamma sub band signals (30-50Hz) are extracted using an elliptic band pass filter of sixth order to extract the visually evoked potentials (VEP) signals. Prior to filtering, thresholds of 100v are applied to the electroencephalogram (EEG) recordings in order to remove eye blink artefact. The power spectral densities (PSD's) of the gamma band are calculated using Periodogram and the gamma band spectral entropies are determined. These spectral entropy coefficients in the gamma band are used as features to classify the control subjects from their alcoholic counterparts using multilayer perceptron-back propagation (MLP-BP) and probabilistic neural network(PNN) classifiers. From the experimental study, it can be concluded that the PNN classifier performs better with a classification accuracy of 99% (for a spread factor of #60; 1) than MLP classifier.