Multichannel blind speech source separation with a disjoint constraint source model

The paper discusses multichannel blind speech source separation, a problem of separating different speech sources without much prior information about the mixing system. The authors focus on improving existing algorithms – Multichannel nonnegative matrix factorization (MNMF) and the independent low-rank matrix analysis (ILRMA) – by considering the sparseness of speech signals. They incorporate a disjoint constraint regularizer into both MNMF and ILRMA, developing two enhanced algorithms: s-MNMF and s-ILRMA. The results show improved separation performance.

Publication date: 4 Jan 2024
Project Page: Not Provided
Paper: https://arxiv.org/pdf/2401.01763