Through-Wall Multi-Human Detection and Localization using Ultra-Wideband Impulse Radar (UWB-IR)

Abstract

Efforts to maximize the ability of Ultra-Wideband Impulse Radar (UWB-IR) technology to detect human presence through walls have progressed for years. This study optimized the human-detection capability of UWB-IR for people counting and localization behind walls through signal-processing techniques. These techniques include Clutter Suppression, Noise Reduction, Filtering, Digital Downconversion (DDC), Frame Differencing, Peak Detection, and Clustering. The system was evaluated across 15 unique test scenarios involving human subjects positioned behind walls made of plywood and concrete. Accuracy was assessed by comparing the estimated number of people and their positions with ground-truth data. The average count accuracy for both plywood panel and concrete wall is relatively high, with an overall percentage accuracy of maximum percentage error of 16%. In contrast, distance estimation is better for plywood panel, with a maximum percentage error of 10.85%. Overall, this study achieved Multi-Human Detection, People Counting, and Human Localization with an average accuracy of at least 92.11% across all delimitations considered.