Automatic filtering techniques for three-dimensional kinematics data using 3D motion capture system

The purpose of this paper is to investigate the performance of three algorithms for automatic filtering of 3D displacement data. The first approach is based on power spectrum analysis of signal with auto regressive modeling approach to detect a signal bandwidth in the frequency domain. The second method uses the autocorrelation of the residual signal between filtered and unfiltered data to separate the signal bandwidth from noise. The third approach uses a singular spectrum analysis to detect the variance of the signal and reject the noise based on the eigenvalue decomposition of the signal. Overall, the highest RMS value of 0.480 m/s² was measured in the X direction for PSA method, whereas the lowest RMS value of 0.162 m/s² was recorded for cluster 2 for SSA method. This represents a gain of 3 in accuracy in estimating higher-order derivatives such as linear acceleration of rigid body motion. SSA method is robust and seems to behave well for different signal combinations