Brain-Computer Interfaces (BCIs) use Event-Related Potentials (ERPs)---which are measurable brain responses to different sensory or motor events---to translate neural activity into actionable features for communication and control. Inside these ERPs, the Readiness Potential (RP) is a very important biomarker to identify motor intentionality and movement preparation. However, extracting these fast neural markers from raw electroencephalographic (EEG) data requires a strong artifact suppression and advanced state-space tracking. This thesis investigates and tries to optimize the BCI pre-processing architecture of BVibes, a specialized signal processing toolbox developed at the University of Essex to which this project contributes. We aim to improve this toolbox by adding a Time-Varying Autoregressive (TVAR) Kalman smoother.