Convolutional Feature Vectors and Support Vector Machine for Animal Sound Classification

Kyungdeuk Ko; Sangwook Park; Hanseok Ko

doi:10.1109/EMBC.2018.8512408

Convolutional Feature Vectors and Support Vector Machine for Animal Sound Classification

Kyungdeuk Ko, Sangwook Park, Hanseok Ko

School of Electrical Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

28 Citations (Scopus)

Abstract

Pattern classification based on deep network outperforms conventional methods in many tasks. However, if the database for training exhibits internal representation that lacks substantial discernibility for different classes, the network is considered that learning is essentially failed. Such failure is evident when the accuracy drops sharply in the experiments performing classification task where the animal sounds are observed similar. To address and remedy the learning problem, this paper proposes a novel approach composed of a combination of multiple CNNs each separately pre-trained for generating midlevel features according to each class and then merged into a combined CNN unit with SVM for overall classification. For experiment, animal sound database that include 3 classes with 102 species is firstly established. From the experimental results using the database, the proposed method is shown to outperform over prominent conventional methods.

Original language	English
Title of host publication	40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC 2018
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	376-379
Number of pages	4
ISBN (Electronic)	9781538636466
DOIs	https://doi.org/10.1109/EMBC.2018.8512408
Publication status	Published - 2018 Oct 26
Event	40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC 2018 - Honolulu, United States Duration: 2018 Jul 18 → 2018 Jul 21

Publication series

Name	Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS
Volume	2018-July
ISSN (Print)	1557-170X

Other

Other	40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC 2018
Country/Territory	United States
City	Honolulu
Period	18/7/18 → 18/7/21

Bibliographical note

Funding Information:
* This work was supported by Korea Environment Industry & Technology Institute(KEITI) through Public Technology Program based on Environmental Policy, funded by Korea Ministry of Environment(MOE) (2017000210001).

Funding Information:
ACKNOWLEDGMENT This work was supported by Korea Environment Industry & Technology Institute(KEITI) through Public Technology Program based on Environmental Policy, funded by Korea Ministry of Environment(MOE) (2017000210001).

Publisher Copyright:
© 2018 IEEE.

ASJC Scopus subject areas

Signal Processing
Biomedical Engineering
Computer Vision and Pattern Recognition
Health Informatics

Access to Document

10.1109/EMBC.2018.8512408

Cite this

Ko, K., Park, S., & Ko, H. (2018). Convolutional Feature Vectors and Support Vector Machine for Animal Sound Classification. In 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC 2018 (pp. 376-379). Article 8512408 (Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS; Vol. 2018-July). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/EMBC.2018.8512408

Convolutional Feature Vectors and Support Vector Machine for Animal Sound Classification. / Ko, Kyungdeuk; Park, Sangwook; Ko, Hanseok.
40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC 2018. Institute of Electrical and Electronics Engineers Inc., 2018. p. 376-379 8512408 (Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS; Vol. 2018-July).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Ko, K, Park, S & Ko, H 2018, Convolutional Feature Vectors and Support Vector Machine for Animal Sound Classification. in 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC 2018., 8512408, Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS, vol. 2018-July, Institute of Electrical and Electronics Engineers Inc., pp. 376-379, 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC 2018, Honolulu, United States, 18/7/18. https://doi.org/10.1109/EMBC.2018.8512408

Ko K, Park S, Ko H. Convolutional Feature Vectors and Support Vector Machine for Animal Sound Classification. In 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC 2018. Institute of Electrical and Electronics Engineers Inc. 2018. p. 376-379. 8512408. (Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS). doi: 10.1109/EMBC.2018.8512408

Ko, Kyungdeuk ; Park, Sangwook ; Ko, Hanseok. / Convolutional Feature Vectors and Support Vector Machine for Animal Sound Classification. 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC 2018. Institute of Electrical and Electronics Engineers Inc., 2018. pp. 376-379 (Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS).

@inproceedings{a80d6787332f4092ab3fc6a876f072da,

title = "Convolutional Feature Vectors and Support Vector Machine for Animal Sound Classification",

abstract = "Pattern classification based on deep network outperforms conventional methods in many tasks. However, if the database for training exhibits internal representation that lacks substantial discernibility for different classes, the network is considered that learning is essentially failed. Such failure is evident when the accuracy drops sharply in the experiments performing classification task where the animal sounds are observed similar. To address and remedy the learning problem, this paper proposes a novel approach composed of a combination of multiple CNNs each separately pre-trained for generating midlevel features according to each class and then merged into a combined CNN unit with SVM for overall classification. For experiment, animal sound database that include 3 classes with 102 species is firstly established. From the experimental results using the database, the proposed method is shown to outperform over prominent conventional methods.",

author = "Kyungdeuk Ko and Sangwook Park and Hanseok Ko",

note = "Funding Information: * This work was supported by Korea Environment Industry & Technology Institute(KEITI) through Public Technology Program based on Environmental Policy, funded by Korea Ministry of Environment(MOE) (2017000210001). Funding Information: ACKNOWLEDGMENT This work was supported by Korea Environment Industry & Technology Institute(KEITI) through Public Technology Program based on Environmental Policy, funded by Korea Ministry of Environment(MOE) (2017000210001). Publisher Copyright: {\textcopyright} 2018 IEEE.; 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC 2018 ; Conference date: 18-07-2018 Through 21-07-2018",

year = "2018",

month = oct,

day = "26",

doi = "10.1109/EMBC.2018.8512408",

language = "English",

series = "Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "376--379",

booktitle = "40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC 2018",

}

TY - GEN

T1 - Convolutional Feature Vectors and Support Vector Machine for Animal Sound Classification

AU - Ko, Kyungdeuk

AU - Park, Sangwook

AU - Ko, Hanseok

N1 - Funding Information: * This work was supported by Korea Environment Industry & Technology Institute(KEITI) through Public Technology Program based on Environmental Policy, funded by Korea Ministry of Environment(MOE) (2017000210001). Funding Information: ACKNOWLEDGMENT This work was supported by Korea Environment Industry & Technology Institute(KEITI) through Public Technology Program based on Environmental Policy, funded by Korea Ministry of Environment(MOE) (2017000210001). Publisher Copyright: © 2018 IEEE.

PY - 2018/10/26

Y1 - 2018/10/26

N2 - Pattern classification based on deep network outperforms conventional methods in many tasks. However, if the database for training exhibits internal representation that lacks substantial discernibility for different classes, the network is considered that learning is essentially failed. Such failure is evident when the accuracy drops sharply in the experiments performing classification task where the animal sounds are observed similar. To address and remedy the learning problem, this paper proposes a novel approach composed of a combination of multiple CNNs each separately pre-trained for generating midlevel features according to each class and then merged into a combined CNN unit with SVM for overall classification. For experiment, animal sound database that include 3 classes with 102 species is firstly established. From the experimental results using the database, the proposed method is shown to outperform over prominent conventional methods.

AB - Pattern classification based on deep network outperforms conventional methods in many tasks. However, if the database for training exhibits internal representation that lacks substantial discernibility for different classes, the network is considered that learning is essentially failed. Such failure is evident when the accuracy drops sharply in the experiments performing classification task where the animal sounds are observed similar. To address and remedy the learning problem, this paper proposes a novel approach composed of a combination of multiple CNNs each separately pre-trained for generating midlevel features according to each class and then merged into a combined CNN unit with SVM for overall classification. For experiment, animal sound database that include 3 classes with 102 species is firstly established. From the experimental results using the database, the proposed method is shown to outperform over prominent conventional methods.

UR - http://www.scopus.com/inward/record.url?scp=85056626324&partnerID=8YFLogxK

U2 - 10.1109/EMBC.2018.8512408

DO - 10.1109/EMBC.2018.8512408

M3 - Conference contribution

C2 - 30440414

AN - SCOPUS:85056626324

T3 - Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS

SP - 376

EP - 379

BT - 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC 2018

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC 2018

Y2 - 18 July 2018 through 21 July 2018

ER -

Convolutional Feature Vectors and Support Vector Machine for Animal Sound Classification

Abstract

Publication series

Other

Bibliographical note

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this