Locating Motionless Human in Through-Wall NLOS Scenarios With a Commercial mmWave Device

Abstract

Indoor localization technology is essential for enabling location-based services (LBS). Millimeter-wave (mmWave) radar offers the potential for high-precision LBS; however, accurately localizing stationary individuals or those located behind obstacles remains a challenge due to the weakness and susceptibility of crucial echo signals. We propose a feasible scheme for locating motionless human in indoor through-wall nonline-of-sight (NLOS) scenarios, named LoMo, which infers accurate locations of a target from physiological signals. To achieve this goal, we design a coherent accumulation method to improve the target signal-to-noise ratio and strengthen the expression of weak signals. Then, we troubleshoot the coherent multipath problem by introducing an enhanced super-resolution algorithm with Toeplitz reconstruction of the covariance matrix, which enables the accurate separation of location-related signals. Finally, to avoid the ambiguity caused by diverse environmental variables, we present an adaptive clustering method for mmWave point clouds that considers the spatial features of the scene, allowing the obtaining of real location information without intensive manual effort. To validate the feasibility of LoMo, we run extensive experiments covering human diversity, orientations, and obstacle types. Experimental results show that LoMo achieves a localization error within 17 cm, and a correct identification rate of 96.1% within ±30 cm under through-wall NLOS conditions, outperforming state-of-the-art methods in comparison.