An Improved Device-free Localization via Sparse Coding with Log Thresholding Algorithm

As an emerging technology, Device-free localization (DFL) can localize and track targets by using the space signal perturbation induced by the obstacle in wireless sensor network without carrying any devices. However, such DFL systems still face challenges in robustness and efficiency with low signal-to-noise ratio (SNR) in more complex environments. To address the issues, we proposed a subspace sparse coding iterative log thresholding algorithm (SSC-ILT) for DFL. The log thresholding regularizer is applied to the objective function as the penalty to derive sparse solutions. To meet the requirement of real-time performance, the dimensionality of dictionaries and observed vectors constructed by received signal strength (RSS) measurements can be reduced with principal component analysis (PCA), and the responding coefficient vector is computed via sparse coding with log thresholding operator. By projecting the index of the coefficient vectors to the coordinate of the target in low-dimensional subspace, the location can be estimated. Numerical experiments verified the performance of the proposed algorithm.