A Preliminary Study on Environmental Sound Classification Leveraging Large-Scale Pretrained Model and Semi-Supervised Learning

You Sheng Tsao; Tien Hong Lo; Jiun Ting Li; Shi Yan Weng; Berlin Chen

A Preliminary Study on Environmental Sound Classification Leveraging Large-Scale Pretrained Model and Semi-Supervised Learning

You Sheng Tsao, Tien Hong Lo, Jiun Ting Li, Shi Yan Weng, Berlin Chen

Department of Computer Science and Information Engineering

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

Abstract

With the widespread commercialization of smart devices, research on environmental sound classification has gained more and more attention in recent years. In this paper, we set out to make effective use of large-scale audio pretrained model and semi-supervised model training paradigm for environmental sound classification. To this end, an environmental sound classification method is first put forward, whose component model is built on top a large-scale audio pretrained model. Further, to simulate a low-resource sound classification setting where only limited supervised examples are made available, we instantiate the notion of transfer learning with a recently proposed training algorithm (namely, FixMatch) and a data augmentation method (namely, SpecAugment) to achieve the goal of semi-supervised model training. Experiments conducted on benchmark dataset UrbanSound8K reveal that our classification method can lead to an accuracy improvement of 2.4% in relation to a current baseline method.

Original language	English
Title of host publication	ROCLING 2021 - Proceedings of the 33rd Conference on Computational Linguistics and Speech Processing
Editors	Lung-Hao Lee, Chia-Hui Chang, Kuan-Yu Chen
Publisher	The Association for Computational Linguistics and Chinese Language Processing (ACLCLP)
Pages	103-110
Number of pages	8
ISBN (Electronic)	9789869576949
Publication status	Published - 2021
Event	33rd Conference on Computational Linguistics and Speech Processing, ROCLING 2021 - Taoyuan, Taiwan Duration: 2021 Oct 15 → 2021 Oct 16

Publication series

Name	ROCLING 2021 - Proceedings of the 33rd Conference on Computational Linguistics and Speech Processing

Conference

Conference	33rd Conference on Computational Linguistics and Speech Processing, ROCLING 2021
Country/Territory	Taiwan
City	Taoyuan
Period	2021/10/15 → 2021/10/16

Keywords

Environmental Sound Classification
Semi-supervised learning
Transfer learning

ASJC Scopus subject areas

Language and Linguistics
Linguistics and Language
Speech and Hearing

Cite this

Tsao, Y. S., Lo, T. H., Li, J. T., Weng, S. Y., & Chen, B. (2021). A Preliminary Study on Environmental Sound Classification Leveraging Large-Scale Pretrained Model and Semi-Supervised Learning. In L.-H. Lee, C.-H. Chang, & K.-Y. Chen (Eds.), ROCLING 2021 - Proceedings of the 33rd Conference on Computational Linguistics and Speech Processing (pp. 103-110). (ROCLING 2021 - Proceedings of the 33rd Conference on Computational Linguistics and Speech Processing). The Association for Computational Linguistics and Chinese Language Processing (ACLCLP).

A Preliminary Study on Environmental Sound Classification Leveraging Large-Scale Pretrained Model and Semi-Supervised Learning. / Tsao, You Sheng; Lo, Tien Hong; Li, Jiun Ting et al.
ROCLING 2021 - Proceedings of the 33rd Conference on Computational Linguistics and Speech Processing. ed. / Lung-Hao Lee; Chia-Hui Chang; Kuan-Yu Chen. The Association for Computational Linguistics and Chinese Language Processing (ACLCLP), 2021. p. 103-110 (ROCLING 2021 - Proceedings of the 33rd Conference on Computational Linguistics and Speech Processing).

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

Tsao, YS, Lo, TH, Li, JT, Weng, SY & Chen, B 2021, A Preliminary Study on Environmental Sound Classification Leveraging Large-Scale Pretrained Model and Semi-Supervised Learning. in L-H Lee, C-H Chang & K-Y Chen (eds), ROCLING 2021 - Proceedings of the 33rd Conference on Computational Linguistics and Speech Processing. ROCLING 2021 - Proceedings of the 33rd Conference on Computational Linguistics and Speech Processing, The Association for Computational Linguistics and Chinese Language Processing (ACLCLP), pp. 103-110, 33rd Conference on Computational Linguistics and Speech Processing, ROCLING 2021, Taoyuan, Taiwan, 2021/10/15.

Tsao YS, Lo TH, Li JT, Weng SY, Chen B. A Preliminary Study on Environmental Sound Classification Leveraging Large-Scale Pretrained Model and Semi-Supervised Learning. In Lee LH, Chang CH, Chen KY, editors, ROCLING 2021 - Proceedings of the 33rd Conference on Computational Linguistics and Speech Processing. The Association for Computational Linguistics and Chinese Language Processing (ACLCLP). 2021. p. 103-110. (ROCLING 2021 - Proceedings of the 33rd Conference on Computational Linguistics and Speech Processing).

Tsao, You Sheng ; Lo, Tien Hong ; Li, Jiun Ting et al. / A Preliminary Study on Environmental Sound Classification Leveraging Large-Scale Pretrained Model and Semi-Supervised Learning. ROCLING 2021 - Proceedings of the 33rd Conference on Computational Linguistics and Speech Processing. editor / Lung-Hao Lee ; Chia-Hui Chang ; Kuan-Yu Chen. The Association for Computational Linguistics and Chinese Language Processing (ACLCLP), 2021. pp. 103-110 (ROCLING 2021 - Proceedings of the 33rd Conference on Computational Linguistics and Speech Processing).

@inproceedings{5ccbcd20934846bc8141e541213aea48,

title = "A Preliminary Study on Environmental Sound Classification Leveraging Large-Scale Pretrained Model and Semi-Supervised Learning",

abstract = "With the widespread commercialization of smart devices, research on environmental sound classification has gained more and more attention in recent years. In this paper, we set out to make effective use of large-scale audio pretrained model and semi-supervised model training paradigm for environmental sound classification. To this end, an environmental sound classification method is first put forward, whose component model is built on top a large-scale audio pretrained model. Further, to simulate a low-resource sound classification setting where only limited supervised examples are made available, we instantiate the notion of transfer learning with a recently proposed training algorithm (namely, FixMatch) and a data augmentation method (namely, SpecAugment) to achieve the goal of semi-supervised model training. Experiments conducted on benchmark dataset UrbanSound8K reveal that our classification method can lead to an accuracy improvement of 2.4% in relation to a current baseline method.",

keywords = "Environmental Sound Classification, Semi-supervised learning, Transfer learning",

author = "Tsao, {You Sheng} and Lo, {Tien Hong} and Li, {Jiun Ting} and Weng, {Shi Yan} and Berlin Chen",

note = "Publisher Copyright: {\textcopyright} 2021 ROCLING 2021 - Proceedings of the 33rd Conference on Computational Linguistics and Speech Processing. All rights reserved.; 33rd Conference on Computational Linguistics and Speech Processing, ROCLING 2021 ; Conference date: 15-10-2021 Through 16-10-2021",

year = "2021",

language = "English",

series = "ROCLING 2021 - Proceedings of the 33rd Conference on Computational Linguistics and Speech Processing",

publisher = "The Association for Computational Linguistics and Chinese Language Processing (ACLCLP)",

pages = "103--110",

editor = "Lung-Hao Lee and Chia-Hui Chang and Kuan-Yu Chen",

booktitle = "ROCLING 2021 - Proceedings of the 33rd Conference on Computational Linguistics and Speech Processing",

}

TY - GEN

T1 - A Preliminary Study on Environmental Sound Classification Leveraging Large-Scale Pretrained Model and Semi-Supervised Learning

AU - Tsao, You Sheng

AU - Lo, Tien Hong

AU - Li, Jiun Ting

AU - Weng, Shi Yan

AU - Chen, Berlin

PY - 2021

Y1 - 2021

N2 - With the widespread commercialization of smart devices, research on environmental sound classification has gained more and more attention in recent years. In this paper, we set out to make effective use of large-scale audio pretrained model and semi-supervised model training paradigm for environmental sound classification. To this end, an environmental sound classification method is first put forward, whose component model is built on top a large-scale audio pretrained model. Further, to simulate a low-resource sound classification setting where only limited supervised examples are made available, we instantiate the notion of transfer learning with a recently proposed training algorithm (namely, FixMatch) and a data augmentation method (namely, SpecAugment) to achieve the goal of semi-supervised model training. Experiments conducted on benchmark dataset UrbanSound8K reveal that our classification method can lead to an accuracy improvement of 2.4% in relation to a current baseline method.

AB - With the widespread commercialization of smart devices, research on environmental sound classification has gained more and more attention in recent years. In this paper, we set out to make effective use of large-scale audio pretrained model and semi-supervised model training paradigm for environmental sound classification. To this end, an environmental sound classification method is first put forward, whose component model is built on top a large-scale audio pretrained model. Further, to simulate a low-resource sound classification setting where only limited supervised examples are made available, we instantiate the notion of transfer learning with a recently proposed training algorithm (namely, FixMatch) and a data augmentation method (namely, SpecAugment) to achieve the goal of semi-supervised model training. Experiments conducted on benchmark dataset UrbanSound8K reveal that our classification method can lead to an accuracy improvement of 2.4% in relation to a current baseline method.

KW - Environmental Sound Classification

KW - Semi-supervised learning

KW - Transfer learning

UR - http://www.scopus.com/inward/record.url?scp=85127442839&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85127442839&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:85127442839

T3 - ROCLING 2021 - Proceedings of the 33rd Conference on Computational Linguistics and Speech Processing

SP - 103

EP - 110

BT - ROCLING 2021 - Proceedings of the 33rd Conference on Computational Linguistics and Speech Processing

A2 - Lee, Lung-Hao

A2 - Chang, Chia-Hui

A2 - Chen, Kuan-Yu

PB - The Association for Computational Linguistics and Chinese Language Processing (ACLCLP)

T2 - 33rd Conference on Computational Linguistics and Speech Processing, ROCLING 2021

Y2 - 15 October 2021 through 16 October 2021

ER -

A Preliminary Study on Environmental Sound Classification Leveraging Large-Scale Pretrained Model and Semi-Supervised Learning

Abstract

Publication series

Conference

Keywords

ASJC Scopus subject areas

Other files and links

Fingerprint

Cite this