New dataset for query-based universal sound separation

Arxiv CS.SD (Sound/Audio ML)

Authors propose a semantically consistent dataset for data-efficient query-based universal sound separation, aiming to isolate specific sources from mixtures. The method reduces residual interference in complex acoustic scenarios.

sound-separation

Arxiv CS.SD (Sound/Audio ML)·Kai LiJintao ChengChang ZengZijun YanHelin WangZixiong SuBo ZhengXiaolin Hu

8 days ago