Improvement of Motor Imagery Classification in Brain-Computer Interface Based on EEG Signals Analysis

Rapid developments in AI applications are a pivotal step for the development of modern technologies, highlighting the importance of brain signal classification as one of the promising areas.This study presents a deep learning-driven one-dimensional convolutional neural networks (1D-CNN) model for improving electroencephalogram (EEG) signals classification.The model is trained on different data balance scenarios to determine their impact on performance, and preprocessing techniques, including normalization and principal component analysis, are applied to reduce dimensions and improve performance.The model achieves high classification accuracy depending on the BNCI Horizon 2020 dataset, with the average performance accuracy for the eight classes reaching 91.01% in training and 90.99% in testing, while for the seven classes, it is 99.6%.The study confirms the effectiveness of using different techniques for preprocessing and deep neural networks in classifying EEG signals, which improves the potential for developing brain-computer interfaces to facilitate communication between humans and technology.