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 2022-05-13 15:00
Composing General Audio Representation by Fusing Multilayer Features of a Pre-trained Model
Daisuke Niizumi, Daiki Takeuchi, Yasunori Ohishi, Noboru Harada, Kunio Kashino (NTT) EA2022-9
Abstract (in Japanese) (See Japanese page) 
(in English) Many application studies rely on audio DNN models pre-trained on a large-scale dataset as essential feature extractors, and they extract features from the last layers.
In this study, we focus on our finding that the middle layer features of existing supervised pre-trained models are more effective than the late layer features for some tasks.
We propose a simple approach to compose features effective for general-purpose applications, which calculates feature vectors along the time frame from middle/late layer outputs, then fuses them.
We evaluated our approach with three existing models and confirmed that our approach brings the performance of these models to a level comparable to that of the state-of-the-art.
Keyword (in Japanese) (See Japanese page) 
(in English) pre-trained model / feature fusion / general-purpose audio representation / / / / /  
Reference Info. IEICE Tech. Rep., vol. 122, no. 20, EA2022-9, pp. 41-45, May 2022.
Paper # EA2022-9 
Date of Issue 2022-05-06 (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 EA2022-9

Conference Information
Committee EA  
Conference Date 2022-05-13 - 2022-05-13 
Place (in Japanese) (See Japanese page) 
Place (in English) Online 
Topics (in Japanese) (See Japanese page) 
Topics (in English)  
Paper Information
Registration To EA 
Conference Code 2022-05-EA 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Composing General Audio Representation by Fusing Multilayer Features of a Pre-trained Model 
Sub Title (in English)  
Keyword(1) pre-trained model  
Keyword(2) feature fusion  
Keyword(3) general-purpose audio representation  
Keyword(4)  
Keyword(5)  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Daisuke Niizumi  
1st Author's Affiliation NTT Corporation (NTT)
2nd Author's Name Daiki Takeuchi  
2nd Author's Affiliation NTT Corporation (NTT)
3rd Author's Name Yasunori Ohishi  
3rd Author's Affiliation NTT Corporation (NTT)
4th Author's Name Noboru Harada  
4th Author's Affiliation NTT Corporation (NTT)
5th Author's Name Kunio Kashino  
5th Author's Affiliation NTT Corporation (NTT)
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 2022-05-13 15:00:00 
Presentation Time 25 minutes 
Registration for EA 
Paper # EA2022-9 
Volume (vol) vol.122 
Number (no) no.20 
Page pp.41-45 
#Pages 5 
Date of Issue 2022-05-06 (EA) 

[Return to Top Page]

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