Combining HMM-based melody extraction and NMF-based soft masking for separating voice and accompaniment from monaural audio.

Yun Wang, Zhijian Ou

Browse the full ICASSP paper archive.