Blind identification of convolutive MIMO systems with 3 sources and 2 sensors (Q1607653)

The goal of blind \(r\)-input \(n\)-output \((n\times r)\) system identification is to identify an unknown system, driven by \(r\) unobservable inputs, based on the \(n\) system outputs. Blind identification of a Multiple-Input Multiple-Output (MIMO) system is of great importance in many applications. Most of the literature on \(n\times r\) MIMO problems refers to the case of \(n\geq r.\) In that case, system identification can lead to recovery of the inputs via deconvolution. This paper considers the problem of blind identification of a convolute MIMO system with more inputs than outputs, and in particular, the 3-inputs 2-outputs case. It is assumed that the inputs are temporally white, non-Gaussian distributed, and spatially independent. Solutions for the scalar MIMO case, within scaling and permutation ambiguities, have been proposed in the past, based on the canonical decomposition of tensors constructed from higher-order cross-cumulants of the system output. Here, the problem in the frequency domain is considered, where, for each frequency a number of tensors based on cross-polyspectra of the output is constructed. These tensors lead to the system frequency response within frequency dependent scaling and permutation ambiguities. Some ways to resolve these ambiguities are proposed, and the system response within a scalar and a linear phase is obtained.