Fig. 2

Schematic illustration of the metrics proposed to track user learning. The between-class distance represents the distance between the means of the EEG features distribution of the two motor imagery classes (i.e., both hands, both feet). The within-class distance is computed separately for the two classes as the distance of the means of the EEG features distribution with respect to the first day of training. In the channels’ domain the two metrics were calculated using the Euclidean distance, while in the Riemann domain we considered the geodesic distance (i.e., the shortest path between feature distributions following the Riemannian manifold \({\mathcal {M}}\))