Independent Low-Rank Matrix Analysis Based on Complex Student's $t$-Distribution for Blind Audio Source Separation

摘要

In this paper, we generalize a source generative model in a state-of-the-artblind source separation (BSS), independent low-rank matrix analysis (ILRMA).ILRMA is a unified method of frequency-domain independent component analysisand nonnegative matrix factorization and can provide better performance foraudio BSS tasks. To further improve the performance and stability of theseparation, we introduce an isotropic complex Student's $t$-distribution as asource generative model, which includes the isotropic complex Gaussiandistribution used in conventional ILRMA. Experiments are conducted using bothmusic and speech BSS tasks, and the results show the validity of the proposedmethod.