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 2025-06-13 13:20
Multi-channel Models Knowledge Distillation for Single-Channel Speech Separation
Daichi Nitsu, Roland Hartanto, Koichi Shinoda (Science Tokyo) SP2025-3
Abstract (in Japanese) (See Japanese page) 
(in English) Speech separation is an essential technology for processing speech in noisy environments, with applications in hearing aids, speech recognition systems, and smart speakers. However, single-channel models cannot utilize spatial in- formation and tend to have lower performance. This study proposes MCKD-SS (Multi-Channel Knowledge Distillation for Single-Channel Speech Separation), which distills spatial information from the intermediate layers of a multi-task model that simultaneously performs speech separation and direction of arrival (DOA) estimation. Experiments on the SMS-WSJ and WHAMR! datasets confirmed performance improvements, demonstrating effectiveness in noisy and reverberant environments, although the effect was limited in clean conditions.
Keyword (in Japanese) (See Japanese page) 
(in English) Speech Separation / Knowledge Distillation / Deep Learning / / / / /  
Reference Info. IEICE Tech. Rep., vol. 125, no. 74, SP2025-3, pp. 10-15, June 2025.
Paper # SP2025-3 
Date of Issue 2025-06-06 (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 SP2025-3

Conference Information
Committee SP IPSJ-SLP IPSJ-MUS  
Conference Date 2025-06-13 - 2025-06-14 
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 2025-06-SP-SLP-MUS 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Multi-channel Models Knowledge Distillation for Single-Channel Speech Separation 
Sub Title (in English)  
Keyword(1) Speech Separation  
Keyword(2) Knowledge Distillation  
Keyword(3) Deep Learning  
Keyword(4)  
Keyword(5)  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Daichi Nitsu  
1st Author's Affiliation Institute of Science Tokyo (Science Tokyo)
2nd Author's Name Roland Hartanto  
2nd Author's Affiliation Institute of Science Tokyo (Science Tokyo)
3rd Author's Name Koichi Shinoda  
3rd Author's Affiliation Institute of Science Tokyo (Science Tokyo)
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 2025-06-13 13:20:00 
Presentation Time 140 minutes 
Registration for SP 
Paper # SP2025-3 
Volume (vol) vol.125 
Number (no) no.74 
Page pp.10-15 
#Pages 6 
Date of Issue 2025-06-06 (SP) 

[Return to Top Page]

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