EEG Signal Classification for Motor Imagery Tasks Using a Deep Stacking Network Optimized by Particle Swarm Optimization

Abstract

This research proposes a Deep Stacked Network (DSN) model optimized with a particle swarm optimization (PSO) algorithm for classifying electroencephalogram (EEG) signals in motor imagery tasks. Experiments based on the BCI Competition II dataset confirm the effectiveness of the proposed method, which achieves an accuracy of up to 89.42% and outperforms traditional deep models such as DNN and DBN. The study demonstrates that the combination of DSN and PSO enables robust and accurate recognition of brain states, which is important for brain-computer interface devices.