Conclusion
It is demonstrated that the presented approach, based on Riemannian geometry to deal with the Cross-Session and Cross-Subject classification in Brain-Computer Interfaces, works and allows to eliminate the calibration time in the system.
It is also concluded that analyzing the improvements due to the affine transformation in the Cross-Session and Cross-Subject classification, it can be observed that although in the original data space the results are poor, the affine transformation allows a better accuracy in the classification process.