Paper Abstract and Keywords |
Presentation |
2022-08-04 15:15
[Invited Talk]
Audio Source Separation Combining Wavelet Transform and Deep Neural Network Tomohiko Nakamura (Univ. Tokyo) EA2022-32 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
Audio source separation is a technique of separating an observed audio signal into individual source signals. The use of deep neural networks (DNNs) has improved separation performance and led to the recent emergence of an end-to-end approach, which directly performs separation in the time domain. One of the representative models of this approach is Wave-U-Net. It successively down-samples features with down-sampling (DS) layers, called decimation layers, and up-samples them up to have the time resolution same as an input signal. From the signal processing perspective, we found two underlying problems of the decimation layers: The decimation layers cause aliasing in the feature domain and discard part of features, which may include useful information for source separation. To solve these problems, we proposed a DS layer based on a discrete wavelet transform (DWT). The proposed layer was drawn from our finding that Wave-U-Net resembles multiresolution analysis in architecture. A DWT is used for DS in multiresolution analysis and has an antialiasing filter and perfect reconstruction property. Thus, we can solve the problems of the decimation layers by using the proposed DWT-based DS layer. In this presentation, we introduce an end-to-end audio source separation method using the proposed DWT-based DS layer and our recent studies toward a fusion of signal processing and deep learning. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Audio source separation / Multiresolution analysis / Deep learning / Discrete wavelet transform / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 122, no. 144, EA2022-32, pp. 25-25, Aug. 2022. |
Paper # |
EA2022-32 |
Date of Issue |
2022-07-28 (EA) |
ISSN |
Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
Download PDF |
EA2022-32 |
Conference Information |
Committee |
EA ASJ-H |
Conference Date |
2022-08-04 - 2022-08-05 |
Place (in Japanese) |
(See Japanese page) |
Place (in English) |
|
Topics (in Japanese) |
(See Japanese page) |
Topics (in English) |
|
Paper Information |
Registration To |
EA |
Conference Code |
2022-08-EA-H |
Language |
Japanese |
Title (in Japanese) |
(See Japanese page) |
Sub Title (in Japanese) |
(See Japanese page) |
Title (in English) |
Audio Source Separation Combining Wavelet Transform and Deep Neural Network |
Sub Title (in English) |
|
Keyword(1) |
Audio source separation |
Keyword(2) |
Multiresolution analysis |
Keyword(3) |
Deep learning |
Keyword(4) |
Discrete wavelet transform |
Keyword(5) |
|
Keyword(6) |
|
Keyword(7) |
|
Keyword(8) |
|
1st Author's Name |
Tomohiko Nakamura |
1st Author's Affiliation |
University of Tokyo (Univ. Tokyo) |
2nd Author's Name |
|
2nd Author's Affiliation |
() |
3rd Author's Name |
|
3rd Author's Affiliation |
() |
4th Author's Name |
|
4th Author's Affiliation |
() |
5th Author's Name |
|
5th Author's Affiliation |
() |
6th Author's Name |
|
6th Author's Affiliation |
() |
7th Author's Name |
|
7th Author's Affiliation |
() |
8th Author's Name |
|
8th Author's Affiliation |
() |
9th Author's Name |
|
9th Author's Affiliation |
() |
10th Author's Name |
|
10th Author's Affiliation |
() |
11th Author's Name |
|
11th Author's Affiliation |
() |
12th Author's Name |
|
12th Author's Affiliation |
() |
13th Author's Name |
|
13th Author's Affiliation |
() |
14th Author's Name |
|
14th Author's Affiliation |
() |
15th Author's Name |
|
15th Author's Affiliation |
() |
16th Author's Name |
|
16th Author's Affiliation |
() |
17th Author's Name |
|
17th Author's Affiliation |
() |
18th Author's Name |
|
18th Author's Affiliation |
() |
19th Author's Name |
|
19th Author's Affiliation |
() |
20th Author's Name |
|
20th Author's Affiliation |
() |
Speaker |
Author-1 |
Date Time |
2022-08-04 15:15:00 |
Presentation Time |
60 minutes |
Registration for |
EA |
Paper # |
EA2022-32 |
Volume (vol) |
vol.122 |
Number (no) |
no.144 |
Page |
p.25 |
#Pages |
1 |
Date of Issue |
2022-07-28 (EA) |
|