Article

Inverted pendulum (IP) models of human walking allow for wearable motion-sensor
based estimations of spatio-temporal gait parameters during unconstrained walking
in daily-life conditions. At present it is unclear to what extent different IP
based estimations yield different results, and reliability and validity have not
been investigated in older persons without a specific medical condition. The aim
of this study was to compare reliability and validity of four different IP based
estimations of mean step length in independent-living older persons. Participants
were assessed twice and walked at different speeds while wearing a tri-axial
accelerometer at the lower back. For all step-length estimators, test-retest
intra-class correlations approached or were above 0.90. Intra-class correlations
with reference step length were above 0.92 with a mean error of 0.0 cm when (1)
multiplying the estimated center-of-mass displacement during a step by an
individual correction factor in a simple IP model, or (2) adding an individual
constant for bipedal stance displacement to the estimated displacement during
single stance in a 2-phase IP model. When applying generic corrections or
constants in all subjects (i.e. multiplication by 1.25, or adding 75% of foot
length), correlations were above 0.75 with a mean error of respectively 2.0 and
1.2 cm. Although the results indicate that an individual adjustment of the IP
models provides better estimations of mean step length, the ease of a generic
adjustment can be favored when merely evaluating intra-individual differences.
Further studies should determine the validity of these IP based estimations for
assessing gait in daily life.
CI - Copyright (c) 2013 Elsevier B.V. All rights reserved.

Langue : ANGLAIS