Skip to Main Content
We address the blind identification and deconvolution of multiple input multiple output (MIMO) linear FIR channels. This is an instance of blind separation of convolutive mixtures. The unknown system is decomposed in two factors. The first factor can be deterministically identified from a finite data set; the second factor is shown to belong to a multiplicative group. This last property allows the implementation of effective equivariant deconvolution algorithms, including the maximum likelihood (ML) estimator.