Vertical ground reaction force estimation from benchmark nonstationary kinematic data

Daniel J. Davis, John H. Challis

Research output: Contribution to journalArticlepeer-review

Abstract

Time-differentiating kinematic signals from optical motion capture amplifies the inherent noise content of those signals. Commonly, biomechanists address this problem by applying a Butterworth filter with the same cutoff frequency to all noisy displacement signals prior to differentiation. Nonstationary signals, those with time-varying frequency content, are widespread in biomechanics (eg, those containing an impact) and may necessitate a different filtering approach. A recently introduced signal filtering approach wherein signals are divided into sections based on their energy content and then Butterworth filtered with section-specific cutoff frequencies improved second derivative estimates in a nonstationary kinematic signal. Utilizing this signal-section filtering approach for estimating running vertical ground reaction forces saw more of the signal’s high-frequency content surrounding heel strike maintained without allowing inappropriate amounts of noise contamination in the remainder of the signal. Thus, this signal-section filtering approach resulted in superior estimates of vertical ground reaction forces compared with approaches that either used the same filter cutoff frequency across the entirety of each signal or across the entirety of all signals. Filtering kinematic signals using this signal-section filtering approach is useful in processing data from tasks containing an impact when accurate signal second derivative estimation is of interest.

Original languageEnglish (US)
Pages (from-to)272-276
Number of pages5
JournalJournal of applied biomechanics
Volume37
Issue number3
DOIs
StatePublished - Jun 2021

All Science Journal Classification (ASJC) codes

  • Biophysics
  • Orthopedics and Sports Medicine
  • Rehabilitation

Fingerprint

Dive into the research topics of 'Vertical ground reaction force estimation from benchmark nonstationary kinematic data'. Together they form a unique fingerprint.

Cite this