Device-Free Localization and Identification Using Sub-GHz Passive Radio Mapping

Stijn Denis,Maarten Weyn,Rafael Berkvens,Abdil Kaya

doi:10.3390/app10186183

Stijn Denis, Maarten Weyn + Show 2 more

Open Access

https://doi.org/10.3390/app10186183

Copy DOI

Journal: Applied Sciences	Publication Date: Sep 5, 2020
Citations: 1	License type: CC BY 4.0

Affiliation: University of Antwerp

Abstract

The research domain of device-free localization (DFL) is centered on the study of localization techniques which do not require targets to wear any kind of device. Passive radio mapping or passive fingerprinting is an example of a training-based DFL technique which uses the impact of a human target on radio frequency (RF) communication between stationary nodes to perform localization. We describe a set of experiments performed in a 42 m2 empty office environment in which we installed a RF network with nodes communicating on the 433 MHz and 868 MHz bands. We attempted to locate a single stationary human target based solely on signal strength measurements and did so for six different participants using two different fingerprinting methods. One method was based on Euclidean distance minimization while the other made use of a naive Bayesian classifier. We investigated the impact of frequency band, number of nodes and target body type on localization accuracy. Results indicated that a root mean square error of 48 cm could be obtained with only four nodes, provided that the data from both frequency bands was combined. Additionally, we investigated the potential of these fingerprinting approaches to distinguish between targets based on body type and perform a rudimentary form of passive identification. Accuracy rates for identification could vary significantly depending on target location, with results ranging from 0.07 to 0.75 in the exact same environment. However, the experiment participant with the lowest height and weight could be distinguished from the other participants in over 90% of cases.

Highlights

The term "device-free localization" (DFL) refers to the use of technologies which enable the automatic localization of human and/or nonhuman targets, without requiring the targets to carry a device or tag
In the previous section we have observed that the choice of participant(s) for both training and evaluation could have a significant impact on the localization accuracy
We experimentally investigated the use of two Received Signal Strength (RSS)-based passive fingerprinting techniques for both localization and identification of stationary human targets: one technique based on Euclidean distance minimization and one based on a naive Bayesian classifier

Summary

Introduction

The term "device-free localization" (DFL) refers to the use of technologies which enable the automatic localization of human and/or nonhuman targets, without requiring the targets to carry a device or tag. While there is a wide variety of localization approaches which fall under the umbrella of DFL, ranging from differential air pressure-based techniques [1] to techniques which make use of floor vibrations [2], the term is often used in the context of Radio Frequency (RF)-based DFL In this approach, the impact of the physical presence of targets on RF-waves within an environment is used to infer location-related information about them. The research domain of device-free localization was formally defined for the first time in 2007 by Youssef et al [3] They considered DFL to consist of primarily three aspects: detection, tracking and identification.

Methods

Results

Conclusion