Second-Order Fused Lasso Algorithm for Radio Tomographic Imaging

Abstract

Radio tomographic imaging (RTI) is one of the device-free localization (DFL) approaches used to identify obstacles in a wireless network by using the attenuation information of radio waves. The RTI system's fusion centre collects the received signal strength (RSS) data from all nodes and uses batch estimation to find the spatial loss field (SLF) due to obstacles. The surrounding small noisy pixels in the SLF vector are eliminated by the least absolute shrinkage and selection operator (lasso), which results in improved sparsity in SLF. First-order fused lasso (FL)-based RTI techniques are used for simultaneous improvements in sparsity and structural details of the SLF. However, first-order methods have slower convergence than the second-order methods. Also, this batch FL-based SLF estimation results in high memory requirements. In this letter, a novel second-order fast FL algorithm is proposed to handle such bottlenecks. This algorithm uses a time- and norm-weighted fused lasso (TNWFL) strategy for updating the weights of $\boldsymbol {l}_{1}$ -norm through the use of second-order parameters. The effectiveness of the TNWFL estimator is verified through the simulations.