Clique-based distributed beamforming for speech enhancement in wireless sensor networks

In this paper, we take a graph-theoretical approach to increase the convergence rate of an earlier proposed distributed delay-and-sum beamformer (DDSB) for speech enhancement. Instead of updating estimates across two neighboring nodes as in the DDSB, the proposed clique-based distributed beamformer (CbDB) updates estimates across two neighboring non-overlapping cliques. Theoretical and experimental analysis shows that the proposed method improves the convergence speed of the DDSB. Moreover, the presented approach is more robust than a reference algorithm that is based on clusters, since cliques generally have a better connectivity than clusters. This is also shown by the experimental results.