Nonlinear estimators of human movement in biomechanical signals: Comparison between extended Kalman filter and unscented Kalman filter

Abstract

In the capture of biomechanical signals, the application of Kalman filters allows a correction of the error caused by disturbances or noise in the system. This article presents a comparison of the performance of an Unscented Kalman Filter and an Extended Kalman Filter with the objective of identifying which of the two has a better performance for use in the area of the analysis of human movement. The results of the analysis of the measurement of flexion and extension of the arm, taken with inertial and magnetic sensors, show a Root Mean-Square Error (RMSE) of 2.43º in the Unscented Filter versus a 3.88º in the Extended one. The analysis was done with angular velocities between 1 and 15 rad/s, ideal for biomechanical applications.