Towards an Upper-Limb Robotic Exoskeleton Commanded by a BCI Based on Motor Imagery

Abstract

This work presents a study of brain-computer interfaces (BCIs) to recognize motor imagery tasks, using the Riemannian covariance matrices to compute spatial features, and Linear Discriminant Analysis (LDA), Regularized Discriminant Analysis (RDA), and Support Vector Machine (SVM) as classifiers. These BCIs were evaluated through paired combination of four motor imagery tasks. The chosen BCI achieved promising results (accuracy (ACC) of 81.40%) using RDA, showing the best performance on Subject 1 (ACC = 94.45%). This study is a first stage to obtain a BCI based on motor imagery, in order to increase the effectiveness of a neuro-rehabilitation system using upper-limb robotic exoskeleton.