Fig. 1

Overview of the BCI implementation and training protocol. a BCI pipeline to classify both hands and both feet motor imagery. First, the raw EEG signals were spatially filtered and their power spectral density (PSD) extracted. During the offline calibration, the most discriminative features were identified through canonical variate analysis (CVA) and used to calibrate the decoder to classify the two mental tasks. The BCI commands were then converted into the proper game commands to control the BrainDriver game. During the online evaluation only, continuous feedback about the decoder outputs were visually provided to the user to foster learning. b Timeline illustrating the pilot training protocol and the approximate day of decoder update from the first contact with the pilot to the day of the Cybathlon 2020 Global Edition. Between the end of the Cybathlon 2019 BCI Series (17/09/2019) to the following training session (15/09/2020), the pilot spent almost one year without using any BCI system