Influence of The Filtering Method in The Kinematic Data Consistency of Biomechanical Systems: A Benchmark Example

The inverse dynamic analysis of biomechanical systems is corrupted by numerous sources of error that reduce its usefulness. The most important errors are the raw displacement differentiation and the kinematic inconsistency induced by skin motion. The first type of error is mainly due to the amplification of high-frequency low-amplitude noise introduced by the motion capture system when the raw displacement signals are differentiated. The second source of error, the skin motion artifact, produces violations of the kinematic constraint equations of the multibody system. This work studies the influence of the filtering method in the kinematic consistency of biomechanical systems. The objective is to compare the obtained results using several classical and advanced filtering and smoothing schemes: Butterworth filter, GCVSPL (splines), singular spectrum Analysis and the Hodrick-Prescott filter and filter parameters. A benchmark example that includes computer generated data of a four-bar mechanism was processed using the filtering-consistency method to study the influence of several parameters. The results show that the most important error is the raw data problem. The kinematic consistency must be imposed on the smoothed data; in fact, the consistency condition does not eliminate the high-frequency low-amplitude noise present in the displacement signals. Another important conclusion is that the kinematic consistency can detect the presence of subfiltering or overfiltering.