People Tracking with UWB Radar Using a Multiple-Hypothesis Tracking of Clusters (MHTC) Method

−57− This paper explores the use of Ultra-Wideband (UWB) radar for tracking multiple moving humans. Because the ability to track human movement is useful for the wide range of security and safety applications, a number of technologies have been pursued for human tracking. Computer vision has limited performance in poor visibility conditions, while the performance of infrared imagers can be temperature dependent. Human LADAR signatures may not highly discriminable from other moving clutter, and LADAR performance degrades in dusty and foggy conditions. UWB radar can provide a complementary technology for detecting and tracking humans, particularly in poor visibility or through-wall conditions, as it is little affected by dust and moisture. While this paper considers the problem of tracking humans based solely on UWB radar signals, UWB radar technology can profitably joined with other human tracking modalities to provide more robust tracking and detection in a wider variety of operating conditions. Compared with RF, microwave, and mm-wave radar 1) , UWB radar provides high-resolution ranging and localization due to the fine temporal resolution afforded by its wide signal bandwidth 3) 4) . However, the complex scattering behavior of UWB waveforms poses additional signal processing and tracking problems. In our previous work, Chang, et. al., developed an Expectation-Maximization Kalman Filter (EMKF) algorithm for UWB radar-based tracking of a fixed number of humans . However, because this prior work assumes a fixed number of targets, it is necessary to develop a Multi-Target Tracking (MTT) solution which allows for changing numbers of targets, false measurements (clutter), and missed detections (temporary occlusions). An abundant MTT literature has considered military radar and computer vision tracking applications 7) 8) . However, the key differentiator of MTT for UWB radar-based tracking versus traditional applications is the multitude of observations (multipath scattering) per target in each scan, due to the short spatial extent of the transmitted UWB signal pulse width . Because the multi-path signals have a cluster like nature, a two-level data association problem: individual scatters must be properly associated with the correct clusters, and the clusters must be associated across radar scans to generate a consistent track of human movement. Wolf recently developed a Multiple-Hypothesis Tracking of Clusters (MHTC) algorithm for sorting and tracking extracellular neural recordings , whose measurements arrive in an analogous cluster-like nature. We develop a variant of Wolf’s algorithm for the UWB radar-based multi-human target tracking problem, which extends our previously developed algorithm to the more realistic case of varying target number. Recently, Lau, Arras, and Burgard 12) have