Distributed Multichannel Wiener Filtering for Wireless Acoustic Sensor Networks

TL;DR

The proposed distributed multichannel Wiener filter (dMWF) is shown to outperform DANSE in terms of objective metrics after short operation times, highlighting the benefit of its iterationless design.

Abstract

In a wireless acoustic sensor network (WASN), devices (i.e., nodes) can collaborate through distributed algorithms to collectively perform audio signal processing tasks. This paper focuses on the distributed estimation of node-specific desired speech signals using network-wide Wiener filtering. The objective is to match the performance of a centralized system that would have access to all microphone signals, while reducing the communication bandwidth usage of the algorithm. Existing solutions, such as the distributed adaptive node-specific signal estimation (DANSE) algorithm, converge towards the multichannel Wiener filter (MWF) which solves a centralized linear minimum mean square error (LMMSE) signal estimation problem. However, they do so iteratively, which can be slow and impractical. Many solutions also assume that all nodes observe the same set of sources of interest, which is often not the case in practice. To overcome these limitations, we propose the distributed multichannel Wiener filter (dMWF) for fully connected WASNs. The dMWF is non-iterative and optimal even when nodes observe different sets of sources. In this algorithm, nodes exchange neighbor-pair-specific, low-dimensional (fused) signals estimating the contribution of sources observed by both nodes in the pair. We formally prove the optimality of dMWF and demonstrate its performance in simulated speech enhancement experiments. The proposed algorithm is shown to outperform DANSE in terms of objective metrics after short operation times, highlighting the benefit of its iterationless design.

Figures (6)

• Figure 1: Example of pos scenario in a $K=9$ nodes wasn. Some sets of sources observed by a single node or a pair of nodes are shown as examples. Symbols are used instead of numbered indices for clear visualization, using diamonds ($\diamond$) for speech sources and squares ($\square$) for noise sources.
• Figure 2: Example of fods scenario in a $K=9$ nodes wasn. Some sets of sources observed by a single node or a pair of nodes are shown as examples. The same symbols as in Fig. \ref{['fig:sigmod_pos']} are used.
• Figure 3: Schematic representation of the dmwf, focusing on node $q$ during the discovery step and node $k$ during the estimation step.
• Figure 4: Average $\mathrm{MSE}_{W}$ over 10 randomly generated scenarios for the dmwf, danse, and rsdanse, in either fods (top) or pos scenarios (bottom).
• Figure 5: Example of a generated acoustic scenario. Initialization at $t=0$ s (left) and layout after $t=60$ s (right). The paths followed by nodes and sources across the 60 s of simulated signal are shown by colored lines. The blue solid lines represent the edges of the absorptive partition wall.