Enhancing Motor-Imagery-Based Brain Computer Interface (BCI) Classification Performance Based on Common Spatial Pattern (CSP) Techniques

Abstract

This study develops an optimized Convolutional Neural Network (CNN) architecture integrated with the Common Spatial Pattern (CSP) method to enhance the classification performance of motor-imagery-based brain–computer interface (BCI) systems. Motor-imagery BCIs utilize brain activity to enable communication and control of external devices and have demonstrated encouraging potential as assistive technologies for individuals with motor limitations. However, the non-stationary and complex characteristics of electroencephalography (EEG) signals make the extraction of pertinent discriminative information challenging, thereby limiting classification accuracy and overall system performance. To address this limitation, the proposed approach combines the CSP algorithm, which effectively extracts discriminative spatial features from multichannel EEG signals, with an optimized CNN framework designed to make improved use of the learned spatial filters. The methodology comprises preprocessing of the acquired EEG signals, application of CSP for extraction of relevant spatial features, and subsequent training and classification using the optimized CNN architecture. The performance of the proposed CSP–CNN strategy was evaluated using publicly accessible benchmark datasets and compared with an existing state-of-the-art deep neural network approach based on DNN-FBCSP. The proposed architecture achieved a classification accuracy of 98.80%, compared with 88.10% obtained using the existing DNN-FBCSP approach, demonstrating a substantial improvement in motor-imagery classification performance. In addition to improved classification accuracy, the developed framework is intended to reduce computational complexity and enhance the practical usability of motor-imagery-based BCI systems. These improvements demonstrate the potential of the optimized CSP–CNN framework to support the development of efficient and reliable assistive technologies for individuals with motor disorders.