Adaptive rational subspace estimation: case of nonwhite additive noise
Inbar Fijalkow, Philippe Loubaton · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 1994
In previous work, time domain algorithms were proposed to adaptively estimate a rational and orthonormal spanning of rational source and noise subspaces using a cascadable lossless structure. However, they apply only if the additive channel noise occurring at the receiver sensor array is a spatially and temporally white multivariate process. We propose a new approach in the case where the `useful' signal is corrupted by an additive spatially and/or temporally non-white noise. The source and noise subspaces are characterized by the mean of fourth order statistics under pragmatic assumptions on the sources and noise distributions. A new adaptive algorithm is deduced and its satisfactory asymptotical convergence properties are proved.