Identifying Heel Contact and Toe-Off Using Forceplate Thresholds with a Range of Digital-Filter Cutoff Frequencies

Abstract

Analysis of human gait requires accurate measurement of foot-ground contact, often determined using either foot-ground reaction force thresholds or kinematic data. This study examined the differences in calculating event times across five vertical force thresholds and validated a vertical acceleration-based algorithm as a measure of heel contact and toe-off. The experiment also revealed the accuracy in determining heel contact and toe-off when raw displacement/time data were smoothed using a range of digital filter cutoff frequencies. Four healthy young participants completed 10 walking trials: 5 at normal speed (1.2 m/s) and 5 at fast speed (1.8 m/s). A 3D optoelectric system was synchronized with a forceplate to measure the times when vertical force exceeded (heel contact) or fell below (toe-off) 10, 20, 30, 40, and 50 N. These were then compared and subsequently used to validate an acceleration-based method for calculating heel contact and toe-off with the displacement/time data filtered across a range of four cutoff frequencies. Linear regression analyses showed that during both normal and fast walking, any force threshold within 0 to 50 N could be used to predict heel-contact time. For estimating toe-off low force thresholds, 10 N or less should be used. When raw data were filtered with the optimal cutoff frequency, the absolute value (AbsDt) of the difference between the forceplate event times obtained using a 10-N threshold and the event times of heel contact and toe-off using the acceleration-based algorithms revealed average AbsDt of 10.0 and 16.5 ms for normal walking, and 7.4 and 13.5 ms for fast walking. Data smoothing with the non-optimal cutoff frequencies influenced the event times computed by the algorithms and produced greater AbsDt values. Optimal data filtering procedures are, therefore, essential for ensuring accurate measures of heel contact and toe-off when using the acceleration-based algorithms.