Discrete wavelet transform based on EEG signal analysis for diagnosing neurological disorder

Abstract

ElectroEncephaloGram (EEG) signal analysis is critical since it is a reliable approach for detecting neurological brain diseases. In this work, the artifacts in the EEG dataset are removed using the Independent Component Analysis (ICA) technique. The EEG dataset was then filtered with a band-pass filter to eliminate noise. In this work using a Discrete Wavelet Transform (DWT) to deconstruct the filtered data, the EEG signal features are recovered. The features are also supplied into four separate classifiers. Five statistical techniques are utilised to extract characteristics from EEG sub bands: Local Binary Pattern (LBP), Standard Deviation (SD), Variance, Kurtosis, and Shannon Entropy (SE). To classify the features related to their classes Linear Discriminant Analysis (LDA), Support Vector Machine (SVM), K-Nearest Neighbour (KNN), and Artificial Neural Network (ANN) are four classifiers that use the features.The overall classification accuracy approaches 90.5% using SVM, 99% using ANN, 87.5% using LDA and 97% using KNN respectively. In this work ANN gives better performance accuracy than other classifers.