Physical activity (PA) has been identified as a key health indicator and has been assessed in multiple studies as the study of cardiovascular diseases, diabetes mellitus, obesity and psychological conditions. For this reason the use of accelerometers to quantify PA is widely accepted in either clinical/laboratory settings as well as in freeliving environments. The use of accelerometers in the measurement of PA is a relatively new technique and, therefore, has been not standardized. The aim of this work is to study the influence of signal processing stages in the estimation of PA while measuring simultaneously in three different sites to provide robust indicators of activity in front of changes in site measurement. With our measurement we show that a proper choice of signal processing steps can improve the agreement among activity indices measured from different sites on the same individual. The results also show that the most suited index is the time above threshold. In this case, the best combination of axes is by applying the quadratic mean, the best filtering of axes is using a cut-off frequency of around 1.5 Hz and a threshold to compute the index of 0.04 g.

Peer Reviewed