Multichannel blind separation and deconvolution of sources with arbitrary distributions

摘要: Blind deconvolution and separation of linearly mixed convolved sources is an important challenging task for numerous applications. While several algorithms have shown promise in these tasks, techniques may fail to separate signal mixtures containing both sub- super-Gaussian distributed sources. In this paper, we present a simple efficient extension family that enables the arbitrary non-Gaussian Our technique monitors statistics each outputs separator using rigorously-derived sufficient criterion stability then selects appropriate nonlinearity channel such local convergence conditions algorithm are satisfied. Extensive simulations show validity efficiency our method blindly extract arbitrary-distributed source signals.