Deep Learning-Based EEG Signal Classification for Enhancing Brain-Computer Interface Accuracy
Abstract
Brain Computer Interface (BCI) systems are one of the revolutionary technologies that are currently offering an actual communication between the human brain and the external device that largely depends on the interpretation of the Electroencephalogram (EEG) signals. The present paper has proposed an EEG classification model using deep learning that would assist in enhancing accuracy and real-time operation of BCI. It used Convolutional Neural Networks (CNN) to identify spatial along with Long Short-Term Memory (LSTM) networks to identify temporal patterns and was assisted with the help of the highly sophisticated preprocessing tools, i.e. band-pass filtering and Independent Component Analysis (ICA). As demonstrated in the experimental findings, the presented CNN-LSTM model had a higher accuracy (92.4) and lower latency (245 ms) than the classical machine learning models (SVM, Random Forest) had. The structure was able to describe complex neural dynamics well resulting in more accurate and reliable command generation in the BCI systems. The results indicate that the suggested architecture has great potential in terms of its applicability in neuroprosthetic control, rehabilitation systems as well as assistive communication technologies, enhancing the development of intelligent neural interface systems.