Novel EEG Classification Based on Hellinger Distance for Seizure Epilepsy Detection

This paper introduces a new classifier based on the Hellinger distance to overcome the challenges encountered by standard classifiers in accurately diagnosing seizure epilepsy using electroencephalogram (EEG) signals. They mainly suffer from poor discriminative capacity and sensitivity towards datasets with class imbalance and inefficiency towards handling high-dimensional datasets. In an attempt to overcome such challenges, we include the Hellinger distance classifier with Particle Swarm Optimization (PSO) in our proposed work. We incorporate this dynamic approach in such a way that features of EEG signals are effectively selected, which increases classifier accuracy and reduces the dataset time and dimensionality. The experimental results show that our approach strongly increases the accuracy of our classifier on the Bonn dataset, up to 96.25%, and even the F1-score of 97.74%, recall of 95.59%, and precision of 100%. These results position our method as an effective tool for academic as well as medical applications. In addition, this approach gives a very precise solution to seizure epilepsy detection in EEG signals.