Source extraction in audio via background learning
Yang Wang, Zhengfang Zhou · Inverse Problems and Imaging · 2013
Source extraction in audio is an important problem in the study of blind source separation(BSS) with many practical applications. It is a challenging problem when the foregroundsources to be extracted are weak compared to the background sources. Traditional techniquesoften do not work in this setting. In this paper we propose a novel technique forextracting foreground sources. This is achieved by an interval of silence for theforeground sources. Using this silence interval one can learn the background information,allowing the removal or suppression of background sources. Very effectiveoptimization schemes areproposed for the case of two sources and two mixtures.