Temporal correlations in human locomotion: Recommendations for sampling rate and foot strike detection

Abstract Human stride intervals exhibit statistically persistent fluctuations over a wide range of timescales. Numerous studies have examined stride interval correlations during overground and treadmill walking. However, the practices and procedures for pre-preprocessing gait data vary considerably. This paper examined how differences in sampling rate and gait event detection affected stride interval correlations measured using detrended fluctuation analysis. Sampling rates ranged from 30–360 Hz. Three common kinematic event detection algorithms were compared to a force-based reference. Lower sampling frequencies failed to capture subtle variations in gait cycle timing, which reduced the strength of stride interval correlations. The kinematic event detection algorithms produced slightly higher, but comparable, estimates of stride interval correlations relative to the force-based reference. Guidelines and considerations for assessing stride interval correlations and the implications of these findings are discussed.