Distributed Adaptive Signal Estimation in Wireless Sensor Networks With Partial Prior Knowledge of the Desired Sources Steering Matrix

• Published in: IEEE transactions on signal and information processing over networks 2021, Vol.7, p.478-492
• Main Authors: Van Rompaey, Robbe, Moonen, Marc
• Format: Article
• Language: English
• Subjects: Adaptive algorithms; Algorithms; Computer simulation; Convergence; Correlation; distributed estimation; Eigenvalues; Estimation; generalized eigenvalue decomposition (GEVD); Information processing; Knowledge engineering; multichannel wiener filter (MWF); Nodes; Noise reduction; Optimization; prior knowledge; Reference signals; Sensors; Speech enhancement; Speech processing; Steering; Target detection; Tracking; Wiener filtering; Wireless networks; Wireless sensor networks; Wireless sensor networks (WSN)
• ISSN: 2373-776X; 2373-776X; 2373-7778
• DOI: 10.1109/TSIPN.2021.3098941

In the past two decades, wireless sensor networks (WSNs) and their applications have been the topic of many studies. Different multi-sensor nodes are used to collect, process and distribute data over wireless links to perform different tasks such as smart detection, target tracking, node localization, etc. In this article, the problem of distributed adaptive estimation of node-specific signals for signal enhancement or noise reduction is addressed. First the centralized rank R generalized eigenvalue decomposition (GEVD) based multichannel Wiener filter (MWF) with prior knowledge for node-specific signal estimation in a WSN is introduced, where (some of) the nodes have partial prior knowledge of the desired sources steering matrix. A distributed adaptive estimation algorithm for a fully-connected WSN is then proposed demonstrating that this MWF can be obtained by letting the nodes work on compressed (i.e. reduced-dimensional) sensor signals compared to the centralized algorithm. The distributed algorithm can be used in applications such as speech enhancement in a wireless acoustic sensor network (WASN), where (some of) the nodes have prior knowledge on the location of the desired speech sources and on their local microphone array geometry or have access to clean noise reference signals. Foundations for a proof of convergence using a Lagrangian framework, are given, since convergence is observed in batch-mode simulations. Finally, numerical simulation results are provided for a speech enhancement scenario.