A Three-State Received Signal Strength Model for Device-free Localization

Abstract: The indoor radio propagation channel is typically modeled as a two-state time-variant process where one of the states represents the channel when the environment is static, whereas the other state characterizes the medium when it is altered by people. In this paper, the aforementioned process is augmented with an additional state. It is shown that the changes in received signal strength are dictated by: i) electronic noise, when a person is not present in the monitored area; ii) reflection, when a person is moving in the close vicinity of the line-of-sight; iii) shadowing, when a person is obstructing the line-of-sight component of the transmitter-receiver pair. Statistical and spatial models for the three states are derived and the models are empirically validated. Based on the models, a simplistic device-free localization application is designed which aims to: first, estimate the temporal state of the channel using a hidden Markov model; second, track a person using a particle filter. The results suggest that the tracking accuracy is enhanced by at least 65% while the link's sensitivity region is increased by 100% or more with respect to empirical models presented in earlier works.