Structure-Aided Blind SIMO System Identification

This project will open to several avenues of research in the area of signal processing, specifically in medical, acoustic signal processing applications and communication. Efficient diagnosis techniques, as well as, efficient source separation are made possible through the use of blind system identification methods. Quite a steady research in this field has been made in the literature for over decades. Its analysis has been mostly based on assuming white Gaussian noise environment. In the case of channel estimation, two methods are regularly opposed: training sequence method, which uses the information coming from a known source signal, and blind method that merely uses the information of the received signals. In light of the second method, we intend to explore new challenges in blind system identification and tracking. This will include estimation of long and sparse channels, identification in an impulsive noise environment and the use of side information such as pre-filtering (pulse shaping, anti-aliasing) in communication channels to improve the identification performance. The latter would be investigated and benchmarked against the Cramer-Rao bound (CRB) for the considered system models. Eventually, suitable blind identification algorithms/methods to improve the overall performance will be devised. This will be followed by a performance evaluation to thoroughly investigate the proposed identification algorithms/methods using analysis and simulations.