Universal Distributed Sensing via Random Projections
Duarte, Marco F.
Baraniuk, Richard G.
linear programming; compressed sensing; Sparsity; greedy algorithms; sensor networks; correlation
This paper develops a new framework for distributed coding and compression in sensor networks based on distributed compressed sensing (DCS). DCS exploits both intra-signal and inter-signal correlations through the concept of joint sparsity; just a few measurements of a jointly sparse signal ensemble contain enough information for reconstruction. DCS is well-suited for sensor network applications, thanks to its simplicity, universality, computational asymmetry, tolerance to quantization and noise, robustness to measurement loss, and scalability. It also requires absolutely no inter- sensor collaboration. We apply our framework to several real world datasets to validate the framework.