Sparsity based Radio Tomographic Imaging using Fused Lasso Regularization

The increase in demand of detecting obstructions in a wireless medium without attaching any device with the target is well facilitated by the Radio Tomographic Imaging (RTI) system. Even though it is a promising technique it is a cumbersome task to get the exact position and shape of an object due to ill-posed nature of RTI system. Thus vital task is to effectively choose a regularization technique that not only enhances sparsity by reducing noise after detection but also preserves edges of the object with its appropriate shape by using a heuristic weight model. RTI facilitates us with an imaging vector indicating the loss fields created by obstacles in the medium having knowledge of received signal strength(RSS) values and a weight model that assigns weight to the attenuated pixels in a wireless network. This paper addresses the above-mentioned problem by using a fused lasso regularization via ADMM. The second part of the paper extends performance of fused lasso regularization by implementing it incrementally using distributed learning. The performance metrics shows that fused lasso regularization not only reduces the noise level by increasing the sparsity but also retains the sharp features of the object.