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

Shoko Araki*, Shoji Makino, Hiroshi Sawada, Ryo Mukai

*この研究の対応する著者

研究成果: Chapter

16 被引用数 (Scopus)

抄録

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.

本文言語English
ホスト出版物のタイトルLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
編集者Carlos G. Puntonet, Alberto Prieto
出版社Springer Verlag
ページ898-905
ページ数8
ISBN(電子版)3540230564, 9783540230564
DOI
出版ステータスPublished - 2004
外部発表はい

出版物シリーズ

名前Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
3195
ISSN(印刷版)0302-9743
ISSN(電子版)1611-3349

ASJC Scopus subject areas

  • 理論的コンピュータサイエンス
  • コンピュータサイエンス一般

フィンガープリント

「Underdetermined blind separation of convolutive mixtures of speech with directivity pattern based mask and ICA」の研究トピックを掘り下げます。これらがまとまってユニークなフィンガープリントを構成します。

引用スタイル