A note on data smoothing for movement analysis: the relevance of a nonlinear method

Abstract

The goal of this experiment was to test a potentially useful nonlinear method for smoothing noisy position data, which often is encountered in the analysis of data. This algorithm (7RY) uses a nonlinear smoothing function and behaves like a low-pass filter, automatically removing aberrant points; it is used prior to differentiation of time series so that usable acceleration information can be obtained. The experimental procedure comprises position data collection along with direct accelerometric data recording. From the position-time data, (a) 7RY and (b) Butterworth algorithms have been used to compute twice-differentiated acceleration curves. The directly recorded acceleration measurements were then compared with the acceleration computed from the original position data. Although the results indicated an overall good fit between the recorded and the calculated acceleration curves, only the nonlinear method led to reliable acceleration curves when aberrant points were present in the position data.