Cut-off Frequency Estimation Methods for Biomechanical Data Filtering

Abstract

Motion capture systems are widely used in biomechanical engineeering to obtain 3D kinematic data of the human body. However, during post processing, one of the main problems in the extracted data is the presence of high frequency noise signal. To eliminate the noise in the signal filtering methodology is applied. In order to eleminate the noise in the signal low-pass digital filters are commanly used. Here we focus on design of the filters in order to find the optimum method in determination cutoff fequency of the filter. Fourier Frequency Transformation and Residue methods are the two methodes that have been investigated in this study. For this purpose, "Shoulder Abduction" motion data from 21 (8 male, 13 female) healthy people were collected with a depth camera to obtain 3D positon signal. The collected position data were transferred to the computer for post processing. Using the position data, shoulder movement angles were calculated. The shoulder angle signal with respect to time was analyzed using Cumulative Power Spectral Density (CPSD) and Residue Analysis techniques to estimate better cutoff frequency. The mean values of cutoff frequencies obtained with Residue Analysis and CPSD were 3.121 +/- 0.6 Hz and 0.469 +/- 0.1 respectively. The Pearson correlation coefficient between the two methods was r = 0.501. As a result, the Residue Analysis method proved to produce optimal results in estimation of cut-off frequency.