Robust LMS-based compressive sensing reconstruction algorithm for noisy wireless sensor networks

Abstract

Wireless sensor networks (WSNs) show immense promise in many applications, such as environmental monitoring and remotely metering. Compressive sensing (CS) is a novel signal processing that has been envisioned as a useful regime to address the energy and scaling constraints in WSNs. CS is able to move the burden of sensory nodes to central cloud/server. However, prevailing CS reconstruction algorithms are vulnerable to noise. In this paper, we exploit the natural noise-tolerance property of least mean square (LMS) adaptive filter and propose a greedy-LMS algorithm for CS reconstruction. When SNR is 48dB, greedy-LMS algorithm achieves 16% and 47% higher successful rate than BPDN and OMP, respectively. In addition, the computational complexity of greedy-LMS is competitive with OMP.