Noisy-target Training: A Training Strategy for DNN-based Speech Enhancement without Clean Speech

Takuya Fujimura, Yuma Koizumi, Kohei Yatabe*, Ryoichi Miyazaki

*この研究の対応する著者

研究成果: Conference contribution

18 被引用数 (Scopus)

抄録

Deep neural network (DNN)-based speech enhancement ordinarily requires clean speech signals as the training target. However, collecting clean signals is very costly because they must be recorded in a studio. This requirement currently restricts the amount of training data for speech enhancement to less than 1/1000 of that of speech recognition which does not need clean signals. Increasing the amount of training data is important for improving the performance, and hence the requirement of clean signals should be relaxed. In this paper, we propose a training strategy that does not require clean signals. The proposed method only utilizes noisy signals for training, which enables us to use a variety of speech signals in the wild. Our experimental results showed that the proposed method can achieve the performance similar to that of a DNN trained with clean signals.

本文言語English
ホスト出版物のタイトル29th European Signal Processing Conference, EUSIPCO 2021 - Proceedings
出版社European Signal Processing Conference, EUSIPCO
ページ436-440
ページ数5
ISBN(電子版)9789082797060
DOI
出版ステータスPublished - 2021
イベント29th European Signal Processing Conference, EUSIPCO 2021 - Dublin, Ireland
継続期間: 2021 8月 232021 8月 27

出版物シリーズ

名前European Signal Processing Conference
2021-August
ISSN(印刷版)2219-5491

Conference

Conference29th European Signal Processing Conference, EUSIPCO 2021
国/地域Ireland
CityDublin
Period21/8/2321/8/27

ASJC Scopus subject areas

  • 信号処理
  • 電子工学および電気工学

フィンガープリント

「Noisy-target Training: A Training Strategy for DNN-based Speech Enhancement without Clean Speech」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