IEICE Technical Committee Submission System
Conference Paper's Information
Online Proceedings
[Sign in]
Tech. Rep. Archives
 Go Top Page Go Previous   [Japanese] / [English] 

Paper Abstract and Keywords
Presentation 2019-08-28 14:40
[Poster Presentation] An investigation on training of WaveNet vocoder in end-to-end text-to-speech
Kazuki Yasuhara, Tomoki Hayashi, Tomoki Toda (Nagoya Univ.) SP2019-14
Abstract (in Japanese) (See Japanese page) 
(in English) In this paper, we investigate the training of WaveNet vocoder in end-to-end text-to-speech. Tacotron 2, which is an end-to-end text-to-speech system, consists of a sequence-to-sequence network to estimate acoustic features from text and a WaveNet vocoder to estimates a waveform from acoustic features. Tacotron 2 can generate a very high-quality voice. However, there still remain many unclear points about how to connect and integrate these two networks. To clarify what contributes to the quality of synthesized speech, we focus on a mismatch of acoustic features between training and synthesis of WaveNet vocoder, and we investigate its impact on the quality of synthesized speech from multiple aspects. First, to reduce the mismatch, we apply fine-tuning to WaveNet vocoder by using the acoustic features estimated with Tacotron 2. Second, to improve the robustness against the mismatch, we apply a convolutional neural network (CNN) over time frames to acoustic features. Also, we consider a combination of the fine-tuning and the CNN. The experimental results using a corpus by a single female speaker demonstrate that 1) the naturalness of synthesized speech is not improved by just applying fine-tuning to WaveNet vocoder, 2) the naturalness of synthesized speech is improved by applying the CNN to the acoustic features, and 3) we can expect improvement of naturalness by combining the fine-tuning and the CNN.
Keyword (in Japanese) (See Japanese page) 
(in English) End-to-end text-to-speech / Tacotron 2 / WaveNet vocoder / training procedure / / / /  
Reference Info. IEICE Tech. Rep., vol. 119, no. 188, SP2019-14, pp. 31-36, Aug. 2019.
Paper # SP2019-14 
Date of Issue 2019-08-21 (SP) 
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 SP2019-14

Conference Information
Committee SP  
Conference Date 2019-08-28 - 2019-08-28 
Place (in Japanese) (See Japanese page) 
Place (in English) Kyoto Univ. 
Topics (in Japanese) (See Japanese page) 
Topics (in English)  
Paper Information
Registration To SP 
Conference Code 2019-08-SP 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) An investigation on training of WaveNet vocoder in end-to-end text-to-speech 
Sub Title (in English)  
Keyword(1) End-to-end text-to-speech  
Keyword(2) Tacotron 2  
Keyword(3) WaveNet vocoder  
Keyword(4) training procedure  
Keyword(5)  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Kazuki Yasuhara  
1st Author's Affiliation Nagoya University (Nagoya Univ.)
2nd Author's Name Tomoki Hayashi  
2nd Author's Affiliation Nagoya University (Nagoya Univ.)
3rd Author's Name Tomoki Toda  
3rd Author's Affiliation Nagoya University (Nagoya Univ.)
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 ()
21st Author's Name  
21st Author's Affiliation ()
22nd Author's Name  
22nd Author's Affiliation ()
23rd Author's Name  
23rd Author's Affiliation ()
24th Author's Name  
24th Author's Affiliation ()
25th Author's Name  
25th Author's Affiliation ()
26th Author's Name / /
26th Author's Affiliation ()
()
27th Author's Name / /
27th Author's Affiliation ()
()
28th Author's Name / /
28th Author's Affiliation ()
()
29th Author's Name / /
29th Author's Affiliation ()
()
30th Author's Name / /
30th Author's Affiliation ()
()
31st Author's Name / /
31st Author's Affiliation ()
()
32nd Author's Name / /
32nd Author's Affiliation ()
()
33rd Author's Name / /
33rd Author's Affiliation ()
()
34th Author's Name / /
34th Author's Affiliation ()
()
35th Author's Name / /
35th Author's Affiliation ()
()
36th Author's Name / /
36th Author's Affiliation ()
()
Speaker Author-1 
Date Time 2019-08-28 14:40:00 
Presentation Time 60 minutes 
Registration for SP 
Paper # SP2019-14 
Volume (vol) vol.119 
Number (no) no.188 
Page pp.31-36 
#Pages
Date of Issue 2019-08-21 (SP) 


[Return to Top Page]

[Return to IEICE Web Page]


The Institute of Electronics, Information and Communication Engineers (IEICE), Japan