Robust Automated Step Extraction From Time-Series Contact Force Data Using the PDShoe

This paper presents a method of stride identification, extraction, and analysis of data sets of time-series contact force data for ambulating subjects both with and without Parkinson's disease (PD). This method has been made robust with the use of seeded K-Means clustering, fast Fourier transformation (FFT) spectral analysis, and minimum window size rejection. These methods combine to produce well selected strides of active walking data. We are able to calculate quality of walking measures of stride duration, stance duration (as percent of gait cycle - %GC), swing duration (%GC), time to maximum heel force (%GC), time to maximum toe force (%GC), time spent in heel contact (%GC), and time spent in toe contact (%GC).