Distributed speech signal enhancement based on H ∞-consensus filtering

Haiyang Yu,Rubo Zhang,Guanqun Liu
DOI: https://doi.org/10.13245/j.hust.15S1096
2015-01-01
Abstract:The speech signal enhancement problem in sensor networks was discussed,and a distribu-ted estimation approach was proposed based on the sparsity of signals on frequency domain.By desig-ning a node filter to fuse the measurements over the network,each sensor node got a satisfactory esti-mation.Firstly,a so-called l 1-regularized H ∞ filter was established by means of the classic H ∞ filte-ring theory and a pseudo-measurement equation.A high-pass consensus filter was embed into the l 1-regularized H ∞ filter in information form,and a distributed filtering algorithm was developed,which ensured that all node filters could reach a consensus on the estimates asymptotically and satisfied the prescribed H ∞ performance constraint.Finally,a numerical example was provided to demonstrate ef-fectiveness and applicability of the proposed method.
What problem does this paper attempt to address?