Single Channel ICA for Blind Deconvolution Over Adaptive Networks

Adaptive Networks (AN) comprise a group of agents capable of collecting and processing data, as well as sharing information across the network to collaboratively solve a common task. This work proposes implementing the Single Channel ICA (SCICA) algorithm within the adaptive networks (AN) framework for distributed filter estimation in blind equalization, referred to as AN-SCICA. This algorithm is a technique from a well-studied area of signal processing known as Blind Source Separation, and it has the advantage of being unsupervised while making weak assumptions about the data, which makes it well-suited for broader scenarios of source distributions. Two implementations of the AN-SCICA are proposed: one for a single-task scenario, where the channel is unique across the network; therefore, the agents estimate the same equalization filter, and another for a multitask scenario, where the agents estimate a different yet related set of parameters. For multitasking, the channels will be considered correlated among the agents. To perform equalization in this scenario, this work proposes a two-cascade distributed filter, where the agents share the parameters of the first stage and individually estimate the second stage. The numerical simulations were conducted under various scenarios of source distribution, channel phase characteristics, and considering both synchronous and asynchronous network behaviors, which account for random failures in the network. The results showed improved performance for the distributed SCICA over the AN when compared to its conventional non-cooperative implementation, and the proposed two-filter structure can be successfully applied to the multitask framework.