“…Therefore, machine learning-based BCI models were introduced to reduce individual training session for each BCI use, in which a model has to be calibrated based on the data at the beginning of each session (Ramoser et al, 2000;Blankertz et al, 2002). Recent studies have proposed SMR-based BCI without any session-and subject-specific calibration utilizing the concept of transfer learning (Kang et al, 2009;Li et al, 2010;Lu et al, 2010;Niazi et al, 2013;Kang and Choi, 2014;Fazli et al, 2015;Lotte, 2015;Jayaram et al, 2016;Saha et al, 2017aSaha et al, ,b, 2019Fahimi et al, 2018;He and Wu, 2019).…”