Music Signal Separation Based on Supervised Nonnegative Matrix Factorization with Orthogonality and Maximum-Divergence Penalties

Daichi KITAMURA  Hiroshi SARUWATARI  Kosuke YAGI  Kiyohiro SHIKANO  Yu TAKAHASHI  Kazunobu KONDO  

IEICE TRANSACTIONS on Fundamentals of Electronics, Communications and Computer Sciences   Vol.E97-A    No.5    pp.1113-1118
Publication Date: 2014/05/01
Online ISSN: 1745-1337
DOI: 10.1587/transfun.E97.A.1113
Type of Manuscript: LETTER
Category: Engineering Acoustics
music signal separation,  nonnegative matrix factorization,  supervised method,  

Full Text: PDF(3.8MB)>>
Buy this Article

In this letter, we address monaural source separation based on supervised nonnegative matrix factorization (SNMF) and propose a new penalized SNMF. Conventional SNMF often degrades the separation performance owing to the basis-sharing problem. Our penalized SNMF forces nontarget bases to become different from the target bases, which increases the separated sound quality.