A Deep Learning Architecture for Spatio-Temporal Feature Extraction and Alcoholism Detection

Abstract

Alcoholism is a serious disorder that poses a problem for our society. Detection of alcoholism has no widely accepted standard tests or procedures. An electroencephalogram (EEG) is a method to measure the brain’s electrical activity and can be used to detect alcoholism. These EEG signals are complex and multichannel and hence can be hard to interpret manually. Several previous works have tried to classify a subject as alcoholic or non-alcoholic based on these EEG signals. These works have mostly used machine learning or statistical techniques, along with handcrafted features. Not much work is done on the application of deep learning models for the detection of alcoholism using EEG signals. This paper proposes a novel deep learning architecture that uses a combination of Fast Fourier Transform (FFT), Convolution Neural Network (CNN), Long Short Term Memory (LSTM), and recently proposed Attention mechanism for extracting Spatio-temporal features from multichannel EEG signals. This proposed architecture can classify a subject as alcoholic or control with high accuracy by analyzing EEG signals based on Spatio-temporal features and can be used for automating the task of alcoholism detection.