부산대학교 도서관

Summary: ``Blind source separation (BSS) methods are well suited forthe analysis of multichannel data. In many applications, theobservations are corrupted by an additional structured noise, whichhinders most of the standard BSS techniques. In this article, wepropose a novel robust BSS approach able to jointly unmix the sourcesand separate the source contribution from the structured noise oroutliers. It first builds on a new sparse component modeling thatallows combining both the spectral and morphological/spatial diversityof the sources and the outliers. We introduce the tr-rGMCA algorithm(robust generalized morphological component analysis in transformdomains) to tackle the underlying robust BSS problem. Numericalexperiments highlight the robustness and precision of the proposedmethod in a wide variety of settings, including the full-rank regime.''