Reproduce “MOUSSE: Multilingual Overlapped Speech Separation with Predictive Uncertainty Estimation” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Su Yeon Shim, Yejin Jeon, David Ifeoluwa Adelani, Young-Joo Suh

MOUSSE addresses multilingual speech separation by extending a flow-matching separator with speaker-activity boundary weighting and uncertainty estimates. The estimates guide adaptive sampling, while boundary weighting emphasizes active frames near transitions. On synthetic English–Korean mixtures with two to four speakers, MOUSSE reports 13.18 dB SI-SDRi, 1.58 dB above SR-CorrNet; the result does not establish performance on real conversations or other languages.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.