Distributed Blind Estimation Over Sensor Networks

Abstract

Distributed estimation, where a set of nodes collaboratively estimate some parameters of interest from noisy measurements, has received much attention in both science and engineering. Recent studies mainly focus on distributed non-blind or training-based estimation, that is, a training signal and its desired output are both known to the distributed receivers. However, in some applications, it is physically difficult, if not impossible, to get a training signal in prior. Besides, the use of training signals consumes much channel bandwidth. So, it is more preferable for the receivers to perform distributed estimation without the assistance and expense of training sequences, i.e., distributed blind estimation. In this paper, the problem of distributed blind estimation over sensor networks is considered, and a kind of distributed diffusion generalized Sato algorithm is proposed to design a blind equalizer for channel equalization and source signal estimation. The stability of the proposed method in mean and mean-square senses is analyzed theoretically, and its performance is verified numerically by a series of simulations.