Alcoholic EEG Signal Feature Extraction Based on Relative Wavelet Bispectrum and Bispectrum-Gaussian Using CNN and ANN Classifier

A person who is addicted to alcohol is most likely to have problems related to health and brain function, such as in doing cognitive tasks. Thus, detecting the alcoholic condition is necessary. One of the methods to check whether someone is still addicted to alcohol or not is by looking into their brain signal using an electroencephalograph (EEG). This research compares the performance of two feature extraction methods for EEG signal classification, relative wavelet bispectrum (RWB) and bispectrum-Gaussian; the classifications were done using two different kinds of models, ANN and CNN. According to the experiment's results, the 2D RWB feature and CNN classifier have the highest training accuracy of 99%, while the 1D RWB feature and ANN classifier have the highest testing accuracy of 90%. This resulted in RWB becoming the better EEG feature extraction method than bispectrum-Gaussian. The experiments also suggest that the variation of lag value during the autocorrelation calculation has an impact on the classification accuracy, where for every multiple of two of the lag values, resulting in increasing accuracy by 7.5% on average.