Radio tomographic imaging with feedback-based sparse Bayesian learning

Radio tomographic imaging (RTI) provides an efficient method to realize device-free localization (DFL) which does not require the target to carry any tags or electronic devices. By the measurement of received signal strength (RSS) between node pairs in a wireless sensor network, the attenuation image caused by the target can be reconstructed. Subsequently, the target location can be extracted from the attenuation image. Sparse Bayesian learning (SBL) can be employed for reconstruction because of the sparseness of the attenuation image. However, the fast SBL degrades in reconstruction performances due to the inaccurate estimation on the noise hyper-parameters. To address this, this paper exploits a feedback-based fast SBL framework both for homogeneous-noise and heterogeneous-noise cases. Theoretical modeling and Bayesian inference procedure are given for this feedback-based framework. Finally, RTI experimental results from three different scenarios demonstrate the effectiveness of the proposed scheme.