Feasibility of Using Accelerometers to Detect Human Footsteps for Cadence Estimation on Health Sciences

In recent years, tracking gait parameters of older adults in a non-intrusive way is attacking a lot of attention in the community due to their close correlation with health status. This chapter attempts to investigate the feasibility of using seismic accelerometers to detect human footsteps from floor vibration measurements for cadence estimation. Algorithms building upon high energy peaks in the time-domain acceleration signals were proposed to estimate cadence, and the proposed technique is called the peak acceleration for cadence estimations (PACE). A specialized cadence filter was also developed and implemented to improve the overall cadence estimation from three-step instant cadence estimations. To validate the proposed algorithms, a hallway with four accelerometers placed behind a wall was used as a testbed to measure the structural response generated by the footsteps. A subject walking at three different paces with wireless APDM sensors, the gold standard in health science to measure gait parameters, was used to generate ground-truth data for validation. The results showed that proposed algorithms successfully detect human footsteps with a high degree of accuracy and that the estimated cadence was highly correlated with the cadence measured by the APDM sensors. The specialized filter was able to clean out highly unlikely instant-cadence estimations that substantially improved the overall cadence estimation.