Distributed MAP Estimators for Noise Reduction in Fully Connected Wireless Acoustic Sensor Networks

Several noise reduction algorithms have been proposed for wireless acoustic sensor networks, which consist of spatially distributed nodes that are connected via a wireless link. To decrease the required bandwidth and computational complexity, in this paper we propose two iterative distributed maximum a posteriori (MAP) estimators. In the first scheme, each node sequentially updates its estimate, whereas in the second scheme, all nodes simultaneously update their estimates. Based on simulations in a reverberant room with three nodes, we have compared the noise reduction performance of the proposed distributed MAP estimators with the centralized MAP estimator, where each node has access to all signals, and the local MAP estimator, where each node only has access to its own signals. The simulation results show that the proposed distributed estimators result in a good noise reduction performance, while decreasing the computational complexity compared to the centralized estimator.