Presentation 2022-05-13
Study on noise reduction with a single noisy speech based on Double-DIP
Hien Oonaka, Takuya Fujimura, Ryoichi Miyazaki,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) This paper proposes a new noise reduction method with an untrained Deep Neural Network ( DNN) for a single noisy speech. Image processing based on Deep Image Prior (DIP) has been proposed as a new deep learning framework that does not require pre-training using large amounts of data. DIP focuses on the image generation process in deep learning, and can achieve image processing such as noise reduction for a single degraded image using only an untrained convolutional neural network (CNN). In this study, we first conduct preliminary experiments on noise reduction of speech signals using only untrained CNNs with reference to DIP. Then, we experimentally show that image denoising based on DIP cannot be applied directly to speech signal denoising.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Noise removal / Deep learning / Deep Prior
Paper # EA2022-12
Date of Issue 2022-05-06 (EA)

Conference Information
Committee EA
Conference Date 2022/5/13(1days)
Place (in Japanese) (See Japanese page)
Place (in English) Online
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair Yoshinobu Kajikawa(Kansai Univ.)
Vice Chair Kenichi Furuya(Oita Univ.) / Shoichi Koyama(Univ. of Tokyo)
Secretary Kenichi Furuya(NTT) / Shoichi Koyama(RitsumeikanUniv.)
Assistant Yukou Wakabayashi(Tokyo Metropolitan Univ.) / Tatsuya Komatsu(LINE)

Paper Information
Registration To Technical Committee on Engineering Acoustics
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Study on noise reduction with a single noisy speech based on Double-DIP
Sub Title (in English)
Keyword(1) Noise removal
Keyword(2) Deep learning
Keyword(3) Deep Prior
1st Author's Name Hien Oonaka
1st Author's Affiliation National Institute of Technology, Tokuyama College(NITTC)
2nd Author's Name Takuya Fujimura
2nd Author's Affiliation Nagoya University(Nagoya Univ.)
3rd Author's Name Ryoichi Miyazaki
3rd Author's Affiliation National Institute of Technology, Tokuyama College(NITTC)
Date 2022-05-13
Paper # EA2022-12
Volume (vol) vol.122
Number (no) EA-20
Page pp.pp.54-61(EA),
#Pages 8
Date of Issue 2022-05-06 (EA)