Figure 2.

Implementation of the Kalman Filter algorithm. For each participant of the experiment, the trials are randomly divided into training trials (black) and testing trials (white). Then, the training trials are used for learning of the parameters of the state transition model (matrices A and Q) and the measurement model (matrices H and R). The EMG data from the testing trials and the learned matrices are then used for the prediction of the state vector in both models. Finally, the predictions of the two models are merged via the Kalman filter algorithm and the result of the filter is compared to the actual state vector. The squared correlation coefficient (R²) is used as a measure of efficiency of reconstruction.