Distributed delay and sum beamformer for speech enhancement in wireless sensor networks via randomized gossip

In this paper, we describe a distributed delay and sum beamformer (DDSB) for speech enhancement based on a randomized gossip algorithm. The proposed algorithm operates in a randomly connected wireless sensor network. Without any network topology constraint, the DDSB estimates the desired signal at each node by only exchanging information with its neighbors. Since the DDSB performs only local signal processing, it is robust and scalable for large sensor networks and dynamic environments. We show that the DDSB converges to the optimal estimate of the centralized beamformer. Furthermore, we provide a bound for the worst-case averaging time of the DDSB for the worst connected network. The simulation results validate the theoretical results of the algorithm.