Influence of sampling frequency and number of strides on recurrence quantifiers extracted from gait data

Abstract

In this study, the influence of the sampling frequency and number of strides on recurrence quantifiers extracted from gait data was investigated in order to provide baseline values and preserve the system's non-linear dynamical characteristics expressed by these recurrence quantifiers. Recurrence quantifiers were extracted from a recurrence plot (RP), which required the reconstruction of a high-dimensional state space capable of reproducing the dynamical characteristics of the analyzed system. In this study, the following quantifiers were extracted: rate of recurrence (RR), determinism (DET), average diagonal lines length (AVG), maximum diagonal lines length (MaxL), Shannon entropy (EntD), and measure of trend (TREND). Data collected during treadmill walking were statistically analyzed to compare the distribution characteristics (mean, median, and standard deviation) and the quantifiers' correlation with those obtained from a control time series with an acquisition time corresponding to 150 strides and a 100-Hz sampling frequency, which are common values used in gait studies. It was not possible to reduce the number of strides for the MaxL or TREND. However, for the RR, DET, AVG, and EntD, it was possible to reduce the number of strides by 60% when analyzed together. The minimum sampling frequency required to extract all quantifiers simultaneously was 100 Hz. This potential reduction in the number of strides is appropriate for evaluating fast gait events, with short temporal localization in the RP, by applying the sliding window method to the recurrence plot.