Paper Abstract and Keywords |
Presentation |
2021-03-04 09:00
Anomalous Sound Detection Using a Binary Classification Model Considering Class Centroids Ibuki Kuroyanagi, Tomiki Hayashi, Kazuya Takeda, Tomoki Toda (Nagoya Univ) EA2020-79 SIP2020-110 SP2020-44 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
In an anomalous sound detection system, it is necessary to detect unknown anomalous sounds using only normal sound data.
On the other hand, considering the anomalous sound detection system's operation, it is desirable to build a system that can also use a small amount of anomalous sound data accumulated through the operation.
Therefore, we focus on a binary classification model that can easily use anomalous sound data as well as normal sound data while further using outlier data belonging to the other domain as pseudo-anomalous sound data.
This report presents a new loss function based on metric learning for this binary classification, making it possible to learn the distance relationship from each class centroid in the feature space.
The multi-task learning of classification and metric learning enables us to learn a mapping to the feature space where the within-class variance is minimized and the between-class variance is maximized while keeping normal and anomalous classes linearly separable.
We also investigate the relationship between the amount of anomalous sound data used for training and the detection performance to clarify that the additional use of anomalous sound data is effective for further improving the binary classification model.
Through the experimental evaluation with DCASE~2020 Task~2 dataset, we demonstrate that 1) multi-task learning of binary classification and metric learning is effective, 2) the loss function considering the distance from the class centroids in the feature space is also effective, and 3) using a small amount of anomalous sound data significantly improves the detection performance. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
anomalous sound detection / binary classification / class centriods / semi-supervised learning / metric learning / multi task learning / / |
Reference Info. |
IEICE Tech. Rep., vol. 120, no. 397, EA2020-79, pp. 114-121, March 2021. |
Paper # |
EA2020-79 |
Date of Issue |
2021-02-24 (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 |
EA2020-79 SIP2020-110 SP2020-44 |
Conference Information |
Committee |
EA US SP SIP IPSJ-SLP |
Conference Date |
2021-03-03 - 2021-03-04 |
Place (in Japanese) |
(See Japanese page) |
Place (in English) |
Online |
Topics (in Japanese) |
(See Japanese page) |
Topics (in English) |
Speech, Engineering/Electro Acoustics, Signal Processing, Ultrasonics, and Related Topics |
Paper Information |
Registration To |
EA |
Conference Code |
2021-03-EA-US-SP-SIP-SLP |
Language |
Japanese |
Title (in Japanese) |
(See Japanese page) |
Sub Title (in Japanese) |
(See Japanese page) |
Title (in English) |
Anomalous Sound Detection Using a Binary Classification Model Considering Class Centroids |
Sub Title (in English) |
|
Keyword(1) |
anomalous sound detection |
Keyword(2) |
binary classification |
Keyword(3) |
class centriods |
Keyword(4) |
semi-supervised learning |
Keyword(5) |
metric learning |
Keyword(6) |
multi task learning |
Keyword(7) |
|
Keyword(8) |
|
1st Author's Name |
Ibuki Kuroyanagi |
1st Author's Affiliation |
Nagoya University (Nagoya Univ) |
2nd Author's Name |
Tomiki Hayashi |
2nd Author's Affiliation |
Nagoya University (Nagoya Univ) |
3rd Author's Name |
Kazuya Takeda |
3rd Author's Affiliation |
Nagoya University (Nagoya Univ) |
4th Author's Name |
Tomoki Toda |
4th Author's Affiliation |
Nagoya University (Nagoya Univ) |
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 |
() |
Speaker |
Author-1 |
Date Time |
2021-03-04 09:00:00 |
Presentation Time |
25 minutes |
Registration for |
EA |
Paper # |
EA2020-79, SIP2020-110, SP2020-44 |
Volume (vol) |
vol.120 |
Number (no) |
no.397(EA), no.398(SIP), no.399(SP) |
Page |
pp.114-121 |
#Pages |
8 |
Date of Issue |
2021-02-24 (EA, SIP, SP) |
|