Seizure detection system: A comparative study on features and fusions

Abstract

Human being faces numerous types of neurological disorders. Among them epilepsy is the most frequent after stroke. Several techniques have been developed to identify seizure using EEG signals. The basic contribution of those works can be broadly categorized in three different areas: pre-processing, feature extraction and classification. In this work, we systematically compare different features and their fusions. We have explored how different features and fusions are performing for different cases of seizure classification. We have also investigated how specific combination of features and classifier can outperform others. In addition, we have also observed how information is distributed across different frequency bands for different cases of seizure classifications. Our detailed experimental results illustrate how we can obtain maximum performance by integrating both time and frequency (wavelet) domain features together with specific classifier.