Information: Join today and make your research activities more affordable! Technical workshop participation fees and annual registration fees are available at member rates.
Notice: [Important] Announcement of Changes to Registration Fee Payment and Manuscript Upload Procedures for IEICE Technical Meetings
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 2024-03-01 09:30
SELECTING N-LOWEST SCORES FOR TRAINING MOS PREDICTION MODELS
Yuto Kondo, Hirokazu Kameoka, Kou Tanaka, Takuhiro Kaneko (NTT) EA2023-94 SIP2023-141 SP2023-76
Abstract (in Japanese) (See Japanese page) 
(in English) Automatic speech quality assessment (SQA) is a task to evaluate the quality of speech samples without resorting to time-consuming listener questionnaires.
Attempts have recently been made to train neural-based SQA models to predict the mean opinion score (MOS) of the speech samples produced by text-to-speech or voice conversion systems.
One difficulty in the MOS prediction is that the quality of a (particularly automatically generated) speech sample can vary from segment to segment. Thus, in subjective MOS evaluation, it is up to each listener what segments of the speech sample to focus on to determine the score.
We hypothesize that listeners tend to base their judgments on low-quality segments, and that the variation among listeners in their ratings of each speech sample is primarily due to their mistakenly assigning higher scores by overlooking such segments.
We analyze the VCC2018 and BVCC datasets to support this hypothesis, and propose the use of $N_{rm low}$-MOS, the mean of the $N$-lowest opinion scores, for training MOS predictor models.
Experimental results show that when $N_{rm low}$-MOS was used to train MOSNet, higher LCC and SRCC were obtained than when regular MOS was used, suggesting that $N_{rm low}$-MOS is more likely to reflect subjective speech quality.
Keyword (in Japanese) (See Japanese page) 
(in English) speech quality assessment / mean opinion score / subjective evaluation dataset / training sample selection / MOSNet / / /  
Reference Info. IEICE Tech. Rep., vol. 123, no. 403, SP2023-76, pp. 196-201, Feb. 2024.
Paper # SP2023-76 
Date of Issue 2024-02-22 (EA, SIP, 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 EA2023-94 SIP2023-141 SP2023-76

Conference Information
Committee SIP SP EA IPSJ-SLP  
Conference Date 2024-02-29 - 2024-03-01 
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 2024-02-SIP-SP-EA-SLP 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) SELECTING N-LOWEST SCORES FOR TRAINING MOS PREDICTION MODELS 
Sub Title (in English)  
Keyword(1) speech quality assessment  
Keyword(2) mean opinion score  
Keyword(3) subjective evaluation dataset  
Keyword(4) training sample selection  
Keyword(5) MOSNet  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Yuto Kondo  
1st Author's Affiliation NIPPON TELEGRAPH AND TELEPHONE CORPORATION (NTT)
2nd Author's Name Hirokazu Kameoka  
2nd Author's Affiliation NIPPON TELEGRAPH AND TELEPHONE CORPORATION (NTT)
3rd Author's Name Kou Tanaka  
3rd Author's Affiliation NIPPON TELEGRAPH AND TELEPHONE CORPORATION (NTT)
4th Author's Name Takuhiro Kaneko  
4th Author's Affiliation NIPPON TELEGRAPH AND TELEPHONE CORPORATION (NTT)
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 2024-03-01 09:30:00 
Presentation Time 60 minutes 
Registration for SP 
Paper # EA2023-94, SIP2023-141, SP2023-76 
Volume (vol) vol.123 
Number (no) no.401(EA), no.402(SIP), no.403(SP) 
Page pp.196-201 
#Pages 6 
Date of Issue 2024-02-22 (EA, SIP, SP) 

[Return to Top Page]

[Return to IEICE Web Page]
The Institute of Electronics, Information and Communication Engineers (IEICE), Japan