Wavelet analysis to detect gait events.

Manually detecting gait events by visual inspection of gait data is laborious. Currently, there are no robust techniques available to automate the process. However, detecting gait events is essentially a classification problem; an application for which wavelet analysis, a multiresolution technique, is well suited for. We employ wavelet analysis to classify heel strike- and toe off events using the ground reaction forces that are exerted during walking. We recorded the ground reaction forces for 30 unshod healthy subjects while they were stepping in place on a force platform for 30 s at a self-selected pace. Depending on the pace, each subject completed 14-26 gait cycles. We compared the timing of events detected with the wavelet analysis with the timing of events detected by analyzing the signal time-derivative. On average, the wavelet analysis detected the events 29 ms later. This difference corresponds to 1.2% of the average duration of the gait cycles, which was 2.4 s. Wavelet analysis shows promise for automated detection of gait events.