Blind MIMO System Estimation Based on PARAFAC Decomposition of Higher Order Output Tensors

• Published in: IEEE transactions on signal processing Vol. 54; no. 11; pp. 4156 - 4168
• Main Authors: Acar, T., Yuanning Yu, Petropulu, A.P.
• Format: Journal Article
• Language: English
• Published: New York, NY IEEE 01.11.2006; Institute of Electrical and Electronics Engineers; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Ambiguity; Applied sciences; Blind system identification; Channels; convolutive MIMO; Detection, estimation, filtering, equalization, prediction; Exact sciences and technology; Factorization; Filtering; Frequency domain analysis; Frequency response; Higher order statistics; Image processing; Independent component analysis; Information, signal and communications theory; Mathematical analysis; MIMO; Multiaccess communication; multiple-input multiple-output (MIMO); parallel factorization (PARAFAC); Permutations; Signal and communications theory; Signal processing; Signal, noise; Spectra; Speech processing; System identification; Telecommunications and information theory; Tensile stress; Tensors; MIMO system; Phase unwrapping; Filtering; Image processing; Third order; Linear phase; Frequency domain method; Blind identification; Image restoration; multiple-input multiple-output (MIMO); Factorization; Frequency response; System identification; convolutive MIMO; parallel factorization (PARAFAC); Higher order statistics; Blind system identification
• ISSN: 1053-587X; 1941-0476
• DOI: 10.1109/TSP.2006.879327

We present a novel framework for the identification of a multiple-input multiple-output (MIMO) system driven by white, mutually independent unobservable inputs. Samples of the system frequency response are obtained based on parallel factorization (PARAFAC) of three- or four-way tensors constructed based on, respectively, third- or fourth-order cross spectra of the system outputs. The main difficulties in frequency-domain methods are frequency-dependent permutation and filtering ambiguities. We show that the information available in the higher order spectra allows for the ambiguities to be resolved up to a constant scaling and permutation ambiguities and a linear phase ambiguity. Important features of the proposed approach are that it does not require channel length information, needs no phase unwrapping, and unlike the majority of existing methods, needs no prewhitening of the system outputs