Decoding elbow movement with Kalman Filter using non-invasive EEG

The properties of the Kalman Filter to decode elbow movement from non-invasive EEG are analyzed in this article. A set of configuration parameters using cross-validation are tested in order to find the ones that reduce the estimation error. We found that selecting properly the number of channels and the time step used to configure the signal, it is possible to improve the filter estimation capabilities. As there was a clear incidence of the variations in the recorded data used to train the model, we studied how those alterations affect the estimation precision. Due to the linearity of the filter, major variations in the velocity and acceleration train filters with lower estimation precision than the ones that are built from more homogeneous data.