[Paper Review] A Multi-Scale Spatial Model for RSS-based Device-Free Localization

Ossi Kaltiokallio, Maurizio Bocca+1 more|arXiv (Cornell University)|Feb 24, 2013

Indoor and Outdoor Localization TechnologiesEngineering27 references16 citations

TL;DR

This paper proposes a multi-scale spatial weight model and a measurement model for RSS-based device-free localization (DFL) that accounts for variable spatial impact areas, fade levels, and direction of RSS change. By exploiting channel diversity and modeling human-induced RSS changes more accurately, the system achieves sub-0.30 m localization accuracy in open, cluttered, and through-wall indoor environments without parameter tuning.

ABSTRACT

RSS-based device-free localization (DFL) monitors changes in the received signal strength (RSS) measured by a network of static wireless nodes to locate people without requiring them to carry or wear any electronic device. Current models assume that the spatial impact area, i.e., the area in which a person affects a link's RSS, has constant size. This paper shows that the spatial impact area varies considerably for each link. Data from extensive experiments are used to derive a multi-scale spatial weight model that is a function of the fade level, i.e., the difference between the predicted and measured RSS, and of the direction of RSS change. In addition, a measurement model is proposed which gives a probability of a person locating inside the derived spatial model for each given RSS measurement. A real-time radio tomographic imaging system is described which uses channel diversity and the presented models. Experiments in an open indoor environment, in a typical one-bedroom apartment and in a through-wall scenario are conducted to determine the accuracy of the system. We demonstrate that the new system is capable of localizing and tracking a person with high accuracy (<0.30 m) in all the environments, without the need to change the model parameters.

Motivation & Objective

• To address the limitation of existing DFL models that assume a fixed, link-specific spatial impact area for human-induced RSS changes.
• To model how human presence affects RSS not only on the link line but also at varying distances and directions, depending on multipath conditions.
• To develop a measurement model that estimates the probability of a person being within the modeled spatial impact area based on RSS changes.
• To improve localization accuracy in challenging environments such as cluttered rooms and through-wall scenarios where traditional models fail.
• To demonstrate real-time, high-accuracy DFL without requiring users to carry any electronic devices.

Proposed method

• The multi-scale spatial weight model is derived from extensive experimental data, where the spatial impact area is modeled as a function of the fade level (difference between predicted and measured RSS) and the direction of RSS change (increase or decrease).
• The model accounts for both constructive and destructive multipath interference, showing that RSS can increase or remain unchanged even when a person is not on the direct link line.
• A measurement model is proposed that computes the likelihood of a person’s presence within the spatial impact area based on RSS deviation and fade level.
• The system uses channel diversity by selecting the most anti-fade channels to improve signal-to-noise ratio and localization accuracy.
• A real-time radio tomographic imaging (RTI) system, named msRTI, integrates the new models and processes RSS data in real time to generate accurate localization images.
• The method is validated using a multi-channel, multi-node RF sensor network in three distinct indoor environments: open space, one-bedroom apartment, and through-wall scenario.

Experimental results

Research questions

• RQ1 How does the spatial impact area of human-induced RSS changes vary across different wireless links in indoor environments?
• RQ2 Can the direction of RSS change (increase or decrease) and the fade level be used to model the spatial extent of human-induced signal perturbations more accurately?
• RQ3 Does exploiting channel diversity improve localization accuracy in cluttered and through-wall environments where multipath effects are strong?
• RQ4 To what extent does the proposed multi-scale spatial model reduce localization error compared to conventional RTI methods in challenging propagation environments?
• RQ5 Can the new models achieve consistent high accuracy across diverse indoor environments without requiring re-tuning of parameters?

Key findings

• The proposed multi-scale spatial weight model significantly improves localization accuracy by accounting for variable spatial impact areas that depend on fade level and RSS change direction.
• The system achieves a mean localization error of 0.30 m in through-wall scenarios, outperforming state-of-the-art methods such as flRTI (0.70 m) and cdRTI (1.57 m).
• In open and cluttered environments, the new method (msRTI) achieves sub-0.30 m localization accuracy, demonstrating robustness across diverse propagation conditions.
• The use of channel diversity combined with the new models reduces image noise and improves localization precision, as shown in visual comparisons of estimated propagation field changes.
• The system maintains high accuracy without re-tuning, even in highly challenging environments, indicating strong generalization and adaptability.
• The results confirm that RSS changes can occur far from the link line and that both signal attenuation and amplification are possible, depending on multipath conditions and fade level.