An efficient method for identification of epileptic seizures from EEG signals using Fourier analysis

TLDR: In this article, Fourier decomposition of non-stationary EEG signals has been used for the diagnosis of epilepsy using fast Fourier transform (FFT) algorithm and support vector machine (SVM).

Abstract: Epilepsy is a disease recognized as the chronic neurological dysfunction of the human brain which is described by the sudden and excessive electrical discharges of the brain cells Electroencephalogram (EEG) is a prime tool applied for the diagnosis of epilepsy In this study, a novel and effective approach is introduced to decompose the non-stationary EEG signals using the Fourier decomposition method The concept of position, velocity, and acceleration has been employed on the EEG signals for feature extraction using $$L^p$$ norms computed from Fourier intrinsic band functions (FIBFs) The proposed scheme comprises three main sections In the first section, the EEG signal is decomposed into a finite number of FIBFs In the second stage, the features are extracted from FIBFs and relevant features are selected by using the Kruskal–Wallis test In the last stage, the significant features are passed on to the support vector machine (SVM) classifier By applying 10-fold cross-validation, the proposed method provides better results in comparison to the state-of-the-art methods discussed in the literature, with an average classification accuracy of 9996% and 9994% for classification of EEG signals from the BONN dataset and the CHB-MIT dataset, respectively It can be implemented using the computationally efficient fast Fourier transform (FFT) algorithm