Blind MIMO System Identification

    This work has been supported by the National Science Foundation ( NSF) and also by the Office of Naval Research (ONR)

    The goal of blind system blind identification is to identify an unknown system driven by unobservable inputs, based on the system outputs, and subsequently use the system estimate to recover the input signals (sources). Blind identification of a Multiple-Input Multiple-Output (MIMO) system is of great importance in many applications, since many problems can be formulated as MIMO identification problems. For example, in speech enhancement in the presence of competing speakers, an array of microphones is used to obtain multiple recordings, based on which the signal of interest can be estimated. The microphone outputs can be viewed as the outputs of a MIMO system representing the acoustic environment. MIMO models arise frequently in digital multiuser/multi-access communications systems, digital radio with diversity, multisensor sonar/radar systems. They also arise in biomedical measurements, when recordings of a distributed array of sensors, placed on the skin, are used to pick up signals originating from inside the body.

    Our work considers the convolutive MIMO problem.

    Most of the existing approaches for MIMO system blind identification operate in the time domain. They require a priori knowledge of the order of the mixing system while are sensitive to order mismatch, and their complexity increases rapidly with channel length. We have developed a frequency domain framework that does not require system length information. Common problems with frequency domain approaches are frequency dependent scaling and permutation ambiguities. By exploiting redundancy in the higher-order spectra domain, the proposed framework can effectively deal with the aforementioned ambiguities.

    Sample of our work on this problem can be found in the adjacent Feature section under various Publications

Features