Abstract: This paper presents a human walking speed estimation algorithm using a Doppler radar system for in-home passive gait assessment of elderly adults. The Fast Fourier Transform (FFT) has been a common approach to obtain the gait speed estimation. The proposed algorithm analyzes the received radar signal by counting its zero-crossing periods in the time domain and applies an interquartile range filter to improve its robustness with respect to the gait signature dynamics in the radar data. The proposed algorithm provides better accuracy and does not seem to produce extreme incorrect speed estimates compared to the previously developed FFT method. It has been applied to the Doppler radar system that is deployed in in-home environments to continuously monitor the daily activities of the senior residences and estimate their gait speeds. Improving the accuracy of gait parameters can better facilitate the assessment of the physical conditions of the seniors.
Keywords: Doppler radar, short time Fourier transform, zero-crossing, gait, eldercare