Design an Approach for Epileptic Seizure Detection Using EEG Patterns and Machine Learning Methods

Examining electroencephalogram (EEG) signals plays a crucial role in monitoring brain activity for diagnosing various conditions, with primary applications in detecting epileptic seizures and sleep disorders. Given the diverse range of EEG seizure patterns, there is a significant need to differentiate between normal and seizure data. This study proposes the extraction of time domain statistical and frequency features from EEG signals to enhance classification accuracy. A FIR filter is suggested as a preliminary step to remove artifacts from EEG signals. The research introduces a wavelet-based classifier for extracting EEG frequency features and contributes to the calculation of temporal statistical features using wavelet-based classifications. The study aims to categorize EEG signal patterns into Z, N, F, and S types. Power spectral density (PSD) and peak-to-peak difference (PPD) are utilized as specialized features for classification purposes. Additionally, temporal mean and entropy of EEG signals are employed as statistical features. This expanded feature set can be effectively used in machine learning (ML) based EEG pattern classification, achieving a maximum accuracy of 96.7%.