Underdetermined blind separation of convolutive mixtures of speech with directivity pattern based mask and ICA

Shoko Araki, Shoji Makino, Hiroshi Sawada, Ryo Mukai

Research output: Chapter in Book/Report/Conference proceedingChapter

15 Citations (Scopus)

Abstract

We propose a method for separating N speech signals with M sensors where N > M. Some existing methods employ binary masks to extract the signals, and therefore, the extracted signals contain loud musical noise. To overcome this problem, we propose using a directivity pattern based continuous mask, which masks N - M sources in the observations, and independent component analysis (ICA) to separate the remaining mixtures. We conducted experiments for N = 3 with M = 2 and N = 4 with M = 2, and obtained separated signals with little distortion.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
EditorsCarlos G. Puntonet, Alberto Prieto
PublisherSpringer Verlag
Pages898-905
Number of pages8
ISBN (Electronic)3540230564, 9783540230564
DOIs
Publication statusPublished - 2004
Externally publishedYes

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume3195
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint Dive into the research topics of 'Underdetermined blind separation of convolutive mixtures of speech with directivity pattern based mask and ICA'. Together they form a unique fingerprint.

Cite this