Detecting Human-induced Reflections using RSS of Narrowband Wireless Transceivers

TL;DR

This paper models human-induced reflections in RSS measurements of narrowband wireless transceivers to enable high-probability occupancy detection using multiple frequency channels in indoor environments.

Contribution

It introduces a novel model for human-induced reflections affecting RSS and demonstrates effective occupancy detection with multiple narrowband channels.

Findings

Detection probability exceeds 0.95 with more than eight channels.

False alarm probability is less than 0.01.

Effective detection in a 2m x 2.5m area using a single transceiver pair.

Abstract

Radio frequency sensor networks are becoming increasingly popular as an indoor localization and monitoring technology for gaining unobtrusive situational awareness of the surrounding environment. The localization effort in these networks is built upon the well-established fact that the received signal strength measurements vary due to a person's presence on the line-of-sight of a transmitter-receiver pair. To date, modeling this decrease in received signal strength and utilizing it for localization purposes have received a considerable amount of attention in the research field. However, when the person is in the close vicinity of the line-of-sight but not obstructing it, the signal reflected from the human body is also affecting the received signal strength and can be used for occupancy assessment purposes. In this paper, we first model the effect of human-induced reflections as a function of communication frequency, and then use the model as a basis for energy based occupancy detection. The derived methods are evaluated numerically and the detection probability of the proposed detector is validated with experimental data. The results suggest that when more than eight frequency channels are utilized, presence of a person can be detected using RSS measurements of a single transmit-receive pair with detection probability higher than 0.95 and false alarm probability less than 0.01 in an area of 2 m x 2.5 m. Moreover, the important implications of the studied methods on the available narrowband radio frequency sensor network applications are discussed in detail.