Tracking time-varying mixing system in blind separation
Madalin Ciprian Enescu, Visa Koivunen · 2002
This paper addresses the problem of blind separation from time varying noisy mixtures. The goal is to adapt to changes that occur in the mixing system as well as to estimate if the number of sources change over time. This is achieved by using an adaptive predictor corrector structure and subspace tracker. Changes in the array covariance matrix are used in revealing changes in the mixing system. The performance of the algorithm is studied in simulations.