Active noise control over adaptive distributed networks

Abstract

This paper presents the implementation of Active Noise Control (ANC) systems over a network of distributed acoustic nodes. For this purpose we define a general acoustic node consisting of one or several microphones and one or several loudspeakers together with a unique processor with communication capabilities. ANC systems can use a wide range of adaptive algorithms, but we have considered specifically the Multiple Error Filtered-x Least Mean Square (MEFxLMS), which has been proved to perform very well for ANC systems with multiple microphones and loudspeakers, and centralized processing. We present a new formulation to introduce the distributed version of the MEFxLMS together with an incremental collaborative strategy in the network. We demonstrate that the distributed MEFxLMS exhibits the same performance as the centralized one when there are no communication constraints in the network. Then, we re-formulate the distributed MEFxLMS to include parameters related to its implementation on an acoustic sensor network: latency of the network, computational capacity of the nodes, and trustworthiness of the signals measured at each node. Simulation results in realistic scenarios show the ability of the proposed distributed algorithms to achieve good performance when proper values of these parameters are chosen. (C) 2014 Elsevier B.V. All rights reserved.