Reduction of gait data variability using curve registration

Timing in peak gait values shifts slightly between gait trials. When averag ed, the standard deviation (S.D.) in gait data may increase due to this int er-trial variability unless normalization is carried out beforehand. The ob jective of this study was to determine how curve registration, an alignment technique, can reduce inter-subject variability in gait data without pertu rbing the curve characteristics. Twenty young, healthy men participated in this study each providing a single gait trial. Gait was assessed by means o f a four-camera high-speed video system synchronized to a force plate. A ri gid body three-segment model was used in an inverse dynamic approach to cal culate three-dimensional muscle powers at the hip, knee and ankle. Curve re gistration was applied to each of the 20 gait trials to align the peak powe rs. The mean registered peak powers increased by an average of 0.10 +/- 0.1 3 W/kg with the highest increases in the sagittal plane at push-off. After performing curve registration, the RMS values decreased by 13.6% and the gr eatest reduction occurred at the hip and knee, both in the sagittal plane. No important discontinuities were reported in the first and second derivati ves of the unregistered and registered curves. Curve registration did not h ave much effect on the harmonic content. This would be an appropriate techn ique for application prior to any statistical analysis using able-bodied ga it patterns. (C) 2000 Elsevier Science B.V. All rights reserved.