Rhythm‐based identification of alcohol EEG signals

Abstract

Alcohol is a severe intoxication substance that changes the functionality of brain by disturbing the neuronal process of the central nervous system, which causes mental and behavioural disorders. These disorders can be diagnosed by automatic classification of normal and alcohol electroencephalogram (EEG) signals, because it provides neurophysiology of alcoholic human brain. In this study, EEG rhythms based features are proposed for automatic identification and analysis of alcohol EEG signals. The instantaneous-frequency computed by analytic representation of intrinsic mode functions (IMFs) is used for separation of different frequency ranges known as EEG rhythms. IMFs are obtained by applying empirical mode decomposition on an EEG signal. The variability and complexity of separated EEG rhythms are measured by features namely: mean absolute deviation, inter quartile range, coefficient of variation, entropy, and neg-entropy. The p-value analysis of these features shows that low-frequency (LF)-rhythms based features are statistically significant for identification and neurological analysis of alcohol EEG signals. The LF-rhythms based features are applied as input to an extreme learning machine (ELM) and least squares support vector machine classifiers for classification of normal and alcohol EEG signals. The proposed method with an ELM classifier provides better classification accuracy (97.92%) as compared with other existing methods.