Logistic Regression With Tangent Space-Based Cross-Subject Learning for Enhancing Motor Imagery Classification

Brain-computer interface (BCI) performance is often impacted due to the inherent nonstationarity in recorded EEG signals coupled with a high variability across subjects. This study proposes novel method using logistic regression tangent space-based transfer learning (LR-TSTL) for motor imagery (MI)-based BCI classification problems. The single-trial covariance matrix (CM) features computed from are transformed into Riemannian geometry frame and space by considering lower triangular matrix. These then further classified model improve accuracy. of LR-TSTL tested on healthy subjects’ data set as well stroke patients’ set. As compared existing within-subject approaches proposed gave an equivalent or better terms average accuracy (78.95 ± 11.68%), while applied leave-one-out cross-subject Interestingly, patient significantly ( $p< 0.05$ ) outperformed current benchmark achieving 81.75 6.88%. results show that has potential realize next generation calibration-free technologies enhanced practical usability especially case neurorehabilitative designs patients.