Phase determination during normal running using kinematic data

Algorithms to predict heelstrike and toe-off times during normal running at subject-selected speeds, using only kinematic data, are presented. To assess the accuracy of these algorithms, results are compared with synchronized force platform recordings from ten subjects performing ten trials each. Using a single 180 Hz camera, positioned in the sagittal plane, the average RMS error in predicting heelstrike times is 4.5 ms, whereas the average RMS error in predicting toe-off times is 6.9 ms. Average true errors (negative for an early prediction) are +2.4 ms for heelstrike and +2.8 ms for toe-off, indicating that systematic errors have not occured. The average RMS error in predicting contact time is 7.5 ms, and the average true error in predicting contact time is 0.5 ms. Estimations of event times using these simple algorithms compare favourably with other techniques requiring specialized equipment. It is concluded that the proposed algorithms provide an easy and reliable method of determining event times during normal running at a subject selected pace using only kinematic data and can be implemented with any kinematic data-collection system.