A numerical algorithm for implementing the single-stage maximization criterion for multichannel blind deconvolution

We give an algorithm for the single-stage maximization criterion to attain multichannel blind deconvolution. This criterion determines the coefficients of equalizers for all channels simultaneously. However, the original maximization criterion has many constrains so that it is difficult to directly implement it as a numerical algorithm. By exploiting pre-whitening and lattice representations of paraunitary systems, we can reduce the original maximization problem into a simple constraint-free one and then present a stochastic gradient algorithm.