Publication detail

Musical noise reduction in time-frequency-binary-masking-based blind source separation systems

J. Čermák, S. Araki, H. Sawada, S. Makino

Original Title

Musical noise reduction in time-frequency-binary-masking-based blind source separation systems

Type

conference paper

Language

English

Original Abstract

Blind source separation (BSS) problem consists of estimating N sources from M mixtures without using source and mixing information. In this paper we focus on improving BSS method called time-frequency binary masking (TFBM). TFBM is a versatile approach due to its ability to separate signals in both (over-)determined case (N<=M ) and underdetermined case (N>M). The cost we pay for the versatility is the musical noise which is introduced to the separated signals by the binary masking. We introduce a method, which reduces musical noise and improves separation performance in both (over-)determined and underdetermined cases.

Keywords

Blind source separation, time-frequency binary masking, musical noise

Authors

J. Čermák, S. Araki, H. Sawada, S. Makino

RIV year

2006

Released

27. 9. 2006

Publisher

URE

Location

Praque

ISBN

86269159

Book

16th Czech-German Workshop SPEECH PROCESSING

Pages from

45

Pages to

50

Pages count

6

BibTex

@inproceedings{BUT24879,
  author="Jan {Čermák}",
  title="Musical noise reduction in time-frequency-binary-masking-based blind source separation systems",
  booktitle="16th Czech-German Workshop SPEECH PROCESSING",
  year="2006",
  pages="6",
  publisher="URE",
  address="Praque",
  isbn="86269159"
}