The effects of industry standard averaging and filtering techniques in kinematic gait analysis
Molloy, M., Salazar-Torres, J. J, Kerr, C. F, McDowell, B. C & Cosgrove, AP. (2008). The effects of industry standard averaging and filtering techniques in kinematic gait analysis. Gait and Posture,28(4), 559-562. The Netherlands: Elsevier. Retrieved from https://doi.org/10.1016/j.gaitpost.2008.03.012
Introduction: Conventional methods for determining normative gait patterns consist of filtering marker trajectories prior to processing with subsequent averaging of individual normalized data. This may result in temporal shifts of key parameters and potentially distort normative datasets. Nevertheless, this is the standard method employed by state of the art motion analysis systems. This study compared two different methods of averaging filtered and unfiltered kinematic data. Methods: Forty-eight non-impaired children (22M, 26F, mean age 9.85 years, range 4.2–17 years) underwent three-dimensional gait analysis using a Vicon system (Vicon, Oxford, UK). Gait data were processed with and without the Woltring filtering routine, commonly used to minimize marker trajectory noise. Filtered data were imported into Matlab (MathWorks, Natick, MA) where a representative gait cycle (RGC) for each leg of all participants was selected. Mean and standard deviation values for left and right limbs (n = 96) for filtered and unfiltered data were calculated for seven different parameters within the RGC, without respect to timing. Similar values were obtained using the standard method. The values from the different averaging methods with and without filtering were compared. Results: Differences of up to 2.2° were found between averaging methods and up to 3.31° between filtered and unfiltered data. Discussion: Both the Woltring filtering routine and the standard averaging method cause signal dampening. While a Matlab-based tool may afford greater scope when analysing kinematic gait data, the standard averaging method still returns representative values when utilising data from able-bodied subjects.