Optimizing EEG Artifact Removal in Motor Imagery BCI: Integrating FCIF, FCFBCSP, and Modified DNN for Enhanced Performance

Motor Imagery Brain-Computer Interfaces (MIBCIs) have revolutionized human-computer interaction by giving consumers authority over devices through mental movements. However, EEG signals crucial for decoding imagined movements often suffer from artifacts, challenging the accuracy of MI decoding. This study presents an innovative EEG artifact removal technique tailored for MI BCIs, synergizing Four Class Iterative Filtering (FCIF), Four Class Filter Bank Common Spatial Patterns (FCFBCSP), and a Modified Deep Neural Network (DNN) with Signal-to-Noise Ratio (SNR), Power Spectral Density (PSD), and Spectral Coherence (SC) evaluation. By enhancing specificity in feature extraction and accommodating four MI classes, this approach aims to enhance the reliability and accuracy of MI BCIs.