EEG Classification Using Elliptic Filter and Multilayer Perceptron Based on Gamma Activity Features

Alcohol detection in the community is important to assist in solving the problem caused by alcohol abuse. Machine Learning (ML) in combination with Electroencephalography (EEG) is one alternative solution. Studies on the classification of alcoholic EEG signal has been largely done by transforming the signal to other domains. While the accuracy obtained on the use of time domain feature is less than the transformation method. Another time domain feature type is required to obtain high accuracy. In this study, the time domain features of gamma band were used in the alcoholic and non-alcoholic EEG signal classification. This band activity is a representation of visual stimuli given to humans. Alcoholic detection in this paper consists of three stages: filtering, feature extraction, and classification. Elliptic Highpass filter is used to pass the gamma band EEG signal. The next step, the filtered signal extracted its features only on the time domain. Furthermore, these features are used as input to Multilayer Perceptron (MLP). Based on experiments, features that provide high accuracy are variance and Root Mean Square (RMS) with 96% and 96.08% accuracy respectively.