Combining Optimization And Smoothing Techniques In Human Motion Analysis

Abstract

Obtaining reliable estimates of the degrees of freedom (DOF) of a human body segment by using the trajectories of markers represents a critical problem in human motion analysis. The procedure normally involves the implementation of a smoothing algorithm aimed at reducing noise effects on the measured data and a reconstruction routine, which converts the marker co-ordinates into the elected DOF representation. Three optimized reconstruction and smoothing schemes are introduced and their performances are compared with each other and with those of a more traditional technique. All schemes include an iterative, least-squares DOF reconstruction algorithm and a smoothing routine featuring a self-tuning, zero-phase-shift, 4^-order Butterworth digital filter. In the basic implementation of the filter the optimal cut-off frequency is automatically estimated by analyzing the autocorrelation of the input/output signal residuals. Testing has been carried out on numerically synthesized sequences, added with normally-distributed random signals emulating measurement noise. The results show that filtering the segment DOF previously reconstructed from rough co-ordinates can be as effective as reconstructing the DOF after filtering the marker co-ordinates. In addition, they suggest that the use of an optimization routine provides a visible improvement in DOF reconstruction.