EEG Signal Classification for Alcoholic and Non-Alcoholic Person using Multilevel Wavelet Packet Entropy and Support Vector Machine

Abstract

EEG signal provides information about brain conditions such as brain activity or consciousness level of a person. The consciousness level of a person can also be determined by alcohol. The use of alcohol for a long time can raise specific patterns in EEG signals. Several studies have shown a pattern of differences in EEG signals between alcoholic and non-alcoholic subjects. In this study, EEG signal for alcoholic and non-alcoholic was classified using Multilevel Wavelet Packet Entropy (MWPE) method in the feature extraction stage. MWPE was used to measure the signal complexity at different wavelet decomposition levels. These features are used as Support Vector Machine (SVM) input. The results of the test showed the highest accuracy of 77.8% with quadratic SVM. These results indicated that signal complexity could be used as a differentiator of EEG signals for alcoholic and non-alcoholic persons.