Alco-net: A robust Deep Learning-based framework for Alcoholism detection using EEG signal. This is a novel deep learning framework that is utilized to classify Electroencephalogram (EEG) signals in an attempt to diagnose alcohol intake. Our dataset contains EEG signals collected from both sober and alcoholic subjects, induced by alcohol intake. Some of the preliminary operations involved in data preprocessing were noise removal and normalization of the signals. In this study, to avoid feature engineering, a Convolutional Neural Network (CNN) was developed and trained to extract features as well as classify the data. The proposed model exhibited a mean accuracy level of 99.75%, showcasing that the proposed work achieved high accuracy in differentiating between the EEG signals of sobriety and of alcohol effect. Based on this investigation, it is evident that deep learning methodologies have the capability of advancing the field of biomedical signal processing and affords a strong platform for non-contact alcohol identification. Additionally, our work details a comparison of different deep learning models in order to highlight improved performance metrics of the presented CNN architecture. The marked repeatability and high accuracy of the model render the relevance of this work from the standpoint of healthcare sectors, law enforcement bodies, and occupational health protection services perspectives. Keywords— EEG signals, deep learning, alcohol detection, convolutional neural network, biomedical signal processing
Technight27/Alcoholism_detection
Alco-net: A robust Deep Learning-based framework for Alcoholism detection using EEG signal.
Jupyter NotebookMIT