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

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

DOI：
10.1587/transfun.e97.a.1113
复制
发表时间：
2014-05
期刊：
IEICE Trans. Fundam. Electron. Commun. Comput. Sci.
影响因子：
--
通讯作者：
Daichi Kitamura;H. Saruwatari;Kosuke Yagi;K. Shikano;Yu Takahashi;Kazunobu Kondo
Daichi Kitamura;H. Saruwatari;Kosuke Yagi;K. Shikano;Yu Takahashi;Kazunobu Kondo
中科院分区：
其他
文献类型：
--
作者：
Daichi Kitamura;H. Saruwatari;Kosuke Yagi;K. Shikano;Yu Takahashi;Kazunobu Kondo

文献摘要

被引文献

相似文献

在这封信中，我们解决单声道源分离的基础上监督非负矩阵分解（SNMF），并提出了一个新的惩罚SNMF。传统的SNMF算法由于存在基共享问题，往往会降低分离性能。我们的惩罚SNMF迫使非目标基地成为不同的目标基地，这增加了分离的声音质量。
SUMMARY 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.