Feature extraction and classification technique for chronic alcoholism

In this study, the magnitude and spatial distribution of frequency spectrum in the resting electroencephalogram (EEG) were examined to address the effect of alcoholism on the motor cortex region. The EEG signals for chronic alcoholic subjects were acquired from motor cortex region and Hilbert Huang Transform was applied for Feature Extraction of the EEG signals. The EEG signals were divided into five sub frequency band. It was observed that the extracted feature contained large data dimension, thus it was re duced using Linear Discriminate Analysis. Support Vector Machine has been used for classification of alcoholics in the present study. Maximum classification accuracy (75%) was achieved with the EEG spectral features of maximum spectral coefficient, in the C3 and Cz channel with the combination of beta-1 and beta-2 bands. But more or less high classification accuracy in most of the channels was computed with the EEG features of mean band power. Approximate Entropy, a nonlinear feature was also seen as a dis criminating parameter for alcoholics in the study. Instantaneous Amplitude and Instantaneous Frequency obtained from the Hilbert Spectra