EEG Signal Classification using Memristor-based Reservoir Computing System

Investigations into aberrant brain activities arising from seizure events utilize time series Electroencephalogram (EEG) signals recorded from various regions within the brain. Reservoir Computing (RC) has emerged as an effective machine learning approach for categorizing temporal signals, offering reduced training expenses compared to conventional recurrent neural networks. Recently, memristors have gained significant traction in neuromorphic applications due to their appealing similarity to biological synapses. As a biological application, EEG signal classification is a compatible problem to analyze using memristor-based RC system. This work aims to classify EEG signals (Epileptic vs. Healthy) using a volatile memristor-based reservoir computing system in simulation platform. We have proposed a new RC framework that helps reduce the feature size in the readout layer for classification and achieves 100% accuracy.