Diffusion normalized least mean squares over wireless sensor networks

Recently, distributed adaptive algorithms have been proposed to solve the problem of estimation over distributed networks. In a diffusion protocol, each node in the network function as an individual adaptive filter whose aim is to estimate the parameter of interest through local observations. All the estimates obtained from the nodes are then locally fused with their neighboring estimates in the network. The aim of this work is to improve the signal processing capability of the distributed network in a novel way by applying the diffusion normalized least mean squares (ε-NLMS) algorithm. The simulation results will show that the diffusion ε-NLMS algorithm outperforms its counterpart the diffusion least mean squares (LMS) algorithm for slowly changing environment, where data are expected to show high correlation.