UWB radar-based human target tracking

This paper presents a framework and algorithms for tracking the range of moving humans via a mono-static ultra-wideband (UWB) radar. The approach is based on a specular multi-path model for UWB radar scatters from walking humans. Empirical studies show that multipath time-of-arrival (TOA) can be modeled as a point process whose behavior is governed by a Gamma distribution. Based on this insight, we develop a tracking procedure that combines a Kalman Filter with a point process observation model whose measurements are processed with an Expectation-Maximization (EM) procedure. As a byproduct, the EM procedure solves the multi-target data segmentation and data association problems. We present experimental results in which a monostatic UWB radar tracks both individual and up to two human targets.