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 2023-06-23 13:50
[Poster Presentation] MS-Harmonic-Net++ vs SiFi-GAN: Comparison of fundamental frequency controllable fast neural waveform generative models.
Sota Shimizu (Kobe Univ./NICT), Takuma Okamoto (NICT), Ryoichi Takashima (Kobe Univ.), Yamato Ohtani (NICT), Tetsuya Takiguchi (Kobe Univ.), Tomoki Toda (Nagoya Univ./NICT), Hisashi Kawai (NICT) SP2023-5
Abstract (in Japanese) (See Japanese page) 
(in English) Although Harmonic-Net+ has been proposed as a fundamental frequency (fo) and speech rate (SR) controllable fast neural vocoder with WORLD features, the computational time of WORLD feature extraction itself is slow, and Harmonic-Net+ cannot realize real-time inference when including feature extraction. Then, to realize fo and SR controllable fast neural vocoder including feature extraction, Harmonic-Net++ is proposed with WORLD feature prediction network from mel-spectrogram input. Furthermore, to accelerate the inference speed of Harmonic-Net++ and MS-Harmonic-Net++, which is proposed by introducing multi-stream-based trainable fast upsampling. In this study, we compare MS-Harmonic-Net++ and SiFi-GAN,which is proposed as a high-quality and real-time inference model on CPU by improving on HiFi-GAN as well as Harmonic-Net+, which are fast neural waveform generation models with fo control.
Keyword (in Japanese) (See Japanese page) 
(in English) speech synthesis / neural vocoder / fundamental frequency control / speech rate control / real-time inference / / /  
Reference Info. IEICE Tech. Rep., vol. 123, no. 88, SP2023-5, pp. 20-25, June 2023.
Paper # SP2023-5 
Date of Issue 2023-06-16 (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 SP2023-5

Conference Information
Committee SP IPSJ-MUS IPSJ-SLP  
Conference Date 2023-06-23 - 2023-06-24 
Place (in Japanese) (See Japanese page) 
Place (in English)  
Topics (in Japanese) (See Japanese page) 
Topics (in English)  
Paper Information
Registration To SP 
Conference Code 2023-06-SP-MUS-SLP 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) MS-Harmonic-Net++ vs SiFi-GAN: Comparison of fundamental frequency controllable fast neural waveform generative models. 
Sub Title (in English)  
Keyword(1) speech synthesis  
Keyword(2) neural vocoder  
Keyword(3) fundamental frequency control  
Keyword(4) speech rate control  
Keyword(5) real-time inference  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Sota Shimizu  
1st Author's Affiliation Kobe University/National Institute of Information and Communications Technology (Kobe Univ./NICT)
2nd Author's Name Takuma Okamoto  
2nd Author's Affiliation National Institute of Information and Communications Technology (NICT)
3rd Author's Name Ryoichi Takashima  
3rd Author's Affiliation Kobe University (Kobe Univ.)
4th Author's Name Yamato Ohtani  
4th Author's Affiliation National Institute of Information and Communications Technology (NICT)
5th Author's Name Tetsuya Takiguchi  
5th Author's Affiliation Kobe University (Kobe Univ.)
6th Author's Name Tomoki Toda  
6th Author's Affiliation Nagoya University/National Institute of Information and Communications Technology (Nagoya Univ./NICT)
7th Author's Name Hisashi Kawai  
7th Author's Affiliation National Institute of Information and Communications Technology (NICT)
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 2023-06-23 13:50:00 
Presentation Time 140 minutes 
Registration for SP 
Paper # SP2023-5 
Volume (vol) vol.123 
Number (no) no.88 
Page pp.20-25 
#Pages
Date of Issue 2023-06-16 (SP) 


[Return to Top Page]

[Return to IEICE Web Page]


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