Recreating Periodic Events: Characterizing Footsteps in a Continuous Walking Signal

摘要

Multiple fields present a need to characterize what a footstep response looks like, or detect when footsteps are occurring in a periodic walking signal: localization, classification, and gait analysis among them. At the heart of this problem is a periodic signal encoded with biomechanical information of the person walking. In this work, the periodic nature of gait is used to recreate a template of the floor acceleration response to a single step for a particular person. Underfloor accelerometers positioned to measure vertical acceleration will be used. The floor acceleration response is expected to take the form of a sum of decaying sinusoidal responses, and from this assumed response the expected analytical correlation can be found. Correlation functions of the actual response over the course of multiple steps will be used to exploit the periodic nature of walking to give a clean correlation signal which represents the frequency content of each step. Next, these correlation functions are used to fit coefficients which are in turn used to recreate a template of a single step response. After recreating a template of the footstep event, the template could be used as a direct characterization of the person's gait or to further process the response data such as footstep detection.