Advancing motor imagery EEG classification by quantum feature integration and quantum support vector machines

An advanced approach is presented in the article to classifying motor-imagery (MI) EEG signals. The process begins with identifying the three most significant EEG channels using sequential feature selection (SFS), followed by post-processing using a 5th-order Butterworth bandpass filter. Variational Mode Decomposition (VMD) is then employed to extract features, including statistical metrics and power spectral density (PSD) values across various frequency ranges from the filtered data. Subsequently, a quantum-inspired genetic algorithm (QGA) is applied to optimize the selection of features, exploring more diverse subsets of features to improve the accuracy of the classification. The refined features are then used to train a quantum support vector machine (QSVM), with performance compared to conventional SVM and XGBoost, a tree-based classifier. Our experiments demonstrate that state-of-the-art signal processing techniques combined with quantum algorithms enhance the EEG signal analysis. Specifically, QSVM offers comparable classification accuracy with less data than classical approaches. This innovative method provides new perspectives and opportunities for identifying neurological disorders through advanced instruments, further advancing brain-computer interface (BCI) research.