FBSE-Based Approach for Discriminating Seizure and Normal EEG Signals

Abstract

This letter presents an innovative approach based on Fourier–Bessel series expansion (FBSE) in order to identify seizure and normal electroencephalogram (EEG) signals. The different set of FBSE coefficients are used to separate the five EEG rhythms, namely, delta, theta, alpha, beta, and gamma rhythms. Further, images are generated from the matrices obtained after applying the concept of Euclidean distance on the EEG rhythms. The generated images are employed as features for the classification using convolutional neural network. Notably, our proposed methodology achieves 100% accuracy in distinguishing between seizure and normal EEG signals on the publicly available Bonn University EEG database. This robust performance demonstrates the efficacy of our approach in handling complicated EEG signal patterns. The proposed framework for automated classification of epileptic seizure based on EEG rhythms provides information about the behavior of rhythms during epilepsy. The experimental results on the publicly available Bonn University EEG database show the effectiveness of proposed framework. The performance of the proposed framework is also compared with the other existing frameworks from the literature.