Enhancing the Accuracy of Device-free Localization Using Spectral Properties of the RSS

TL;DR

This paper introduces a spectral analysis approach to improve device-free localization accuracy by leveraging the power spectral density of RSS measurements, enhancing robustness and reducing tracking errors.

Contribution

It proposes using spectral properties of RSS signals to augment localization algorithms, which has not been explored before in this context.

Findings

Tracking accuracy improved by over 50% with spectral augmentation.

System robustness to parameter changes increased.

Validated with both simulations and experimental data.

Abstract

Received signal strength based device-free localization has attracted considerable attention in the research society over the past years to locate and track people who are not carrying any electronic device. Typically, the person is localized using a spatial model that relates the time domain signal strength measurements to the person's position. Alternatively, one could exploit spectral properties of the received signal strength which reflects the rate at which the wireless propagation medium is being altered, an opportunity that has not been exploited in the related literature. In this paper, the power spectral density of the signal strength measurements are related to the person's position and velocity to augment the particle filter based tracking algorithm with an additional measurement. The system performance is evaluated using simulations and validated using experimental data. Compared to a system relying solely on time domain measurements, the results suggest that the robustness to parameter changes is increased while the tracking accuracy is enhanced by 50% or more when 512 particles are used.