Reduction of gait data variability using curve registration

Timing in peak gait values shifts slightly between gait trials. When averaged, the standard deviation (S.D.) in gait data may increase due to this inter-trial variability unless normalization is carried out beforehand. The objective of this study was to determine how curve registration, an alignment technique, can reduce inter-subject variability in gait data without perturbing the curve characteristics. Twenty young, healthy men participated in this study each providing a single gait trial. Gait was assessed by means of a four-camera high-speed video system synchronized to a force plate. A rigid body three-segment model was used in an inverse dynamic approach to calculate three-dimensional muscle powers at the hip, knee and ankle. Curve registration was applied to each of the 20 gait trials to align the peak powers. The mean registered peak powers increased by an average of 0.10 +/- 0.13 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 greatest reduction occurred at the hip and knee, both in the sagittal plane. No important discontinuities were reported in the first and second derivatives of the unregistered and registered curves. Curve registration did not have much effect on the harmonic content. This would be an appropriate technique for application prior to any statistical analysis using able-bodied gait patterns. (C) 2000 Elsevier Science B.V. All rights reserved.