Multichannel Blind Deconvolution of Non minimum Phase Systems Using Information Backpropagation

| W e present a novel method{ lter decomposition approach, for multichannel blind de-convolution of non-minimum phase systems. In 20] we has developed an eecient natural gradient algorithm for causal FIR lters. In this paper we further study the natural gradient method for non-causal lters. We decompose the doubly nite l-ters into a product of two lters, a noncausal FIR lter and a causal FIR lter. The natural gradient algorithm is employed to train the causal FIR lter, and a novel information backpropagation algorithm is developed for training the noncausal FIR lter. Simulations are given to illustrate the eeectiveness and validity o f the algorithm.