Reduced Model Size Deep Convolutional Neural Networks for Small-Footprint Keyword Spotting

Tsung Han Tsai; Xin Hui Lin

doi:10.1109/ICECS53924.2021.9665618

Reduced Model Size Deep Convolutional Neural Networks for Small-Footprint Keyword Spotting

Tsung Han Tsai, Xin Hui Lin

Department of Electrical Engineering

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

2 Scopus citations

Abstract

This paper discussed the application of Densely Connected Convolutional Networks (DenseNet), group convolution, and squeeze-and-excitation Networks (SENet) in keyword spotting tasks. We validated the network using the Google Speech Commands Dataset. Our proposed network has better accuracy than other networks even with less number of parameters and floating-point operations (FLOPs). In addition, we varied the depth and width of the network to build a compact variant network. It also outperforms other compact variants.

Original language	English
Title of host publication	2021 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Proceedings
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9781728182810
DOIs	https://doi.org/10.1109/ICECS53924.2021.9665618
State	Published - 2021
Event	28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Dubai, United Arab Emirates Duration: 28 Nov 2021 → 1 Dec 2021

Publication series

Name	2021 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Proceedings

Conference

Conference	28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021
Country/Territory	United Arab Emirates
City	Dubai
Period	28/11/21 → 1/12/21

Keywords

DenseNet
SENet
deep learning
group convolution
keyword spotting

Access to Document

10.1109/ICECS53924.2021.9665618

Cite this

Tsai, T. H., & Lin, X. H. (2021). Reduced Model Size Deep Convolutional Neural Networks for Small-Footprint Keyword Spotting. In 2021 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Proceedings (2021 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Proceedings). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICECS53924.2021.9665618

Tsai, Tsung Han ; Lin, Xin Hui. / Reduced Model Size Deep Convolutional Neural Networks for Small-Footprint Keyword Spotting. 2021 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2021. (2021 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Proceedings).

@inproceedings{1f6ca5cb56b9496ebe26c055ea647150,

title = "Reduced Model Size Deep Convolutional Neural Networks for Small-Footprint Keyword Spotting",

abstract = "This paper discussed the application of Densely Connected Convolutional Networks (DenseNet), group convolution, and squeeze-and-excitation Networks (SENet) in keyword spotting tasks. We validated the network using the Google Speech Commands Dataset. Our proposed network has better accuracy than other networks even with less number of parameters and floating-point operations (FLOPs). In addition, we varied the depth and width of the network to build a compact variant network. It also outperforms other compact variants.",

keywords = "DenseNet, SENet, deep learning, group convolution, keyword spotting",

author = "Tsai, {Tsung Han} and Lin, {Xin Hui}",

note = "Publisher Copyright: {\textcopyright} 2021 IEEE.; 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 ; Conference date: 28-11-2021 Through 01-12-2021",

year = "2021",

doi = "10.1109/ICECS53924.2021.9665618",

language = "???core.languages.en_GB???",

series = "2021 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Proceedings",

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

booktitle = "2021 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Proceedings",

}

Tsai, TH & Lin, XH 2021, Reduced Model Size Deep Convolutional Neural Networks for Small-Footprint Keyword Spotting. in 2021 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Proceedings. 2021 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Proceedings, Institute of Electrical and Electronics Engineers Inc., 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021, Dubai, United Arab Emirates, 28/11/21. https://doi.org/10.1109/ICECS53924.2021.9665618

Reduced Model Size Deep Convolutional Neural Networks for Small-Footprint Keyword Spotting. / Tsai, Tsung Han; Lin, Xin Hui.
2021 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2021. (2021 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Proceedings).

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

TY - GEN

T1 - Reduced Model Size Deep Convolutional Neural Networks for Small-Footprint Keyword Spotting

AU - Tsai, Tsung Han

AU - Lin, Xin Hui

PY - 2021

Y1 - 2021

N2 - This paper discussed the application of Densely Connected Convolutional Networks (DenseNet), group convolution, and squeeze-and-excitation Networks (SENet) in keyword spotting tasks. We validated the network using the Google Speech Commands Dataset. Our proposed network has better accuracy than other networks even with less number of parameters and floating-point operations (FLOPs). In addition, we varied the depth and width of the network to build a compact variant network. It also outperforms other compact variants.

AB - This paper discussed the application of Densely Connected Convolutional Networks (DenseNet), group convolution, and squeeze-and-excitation Networks (SENet) in keyword spotting tasks. We validated the network using the Google Speech Commands Dataset. Our proposed network has better accuracy than other networks even with less number of parameters and floating-point operations (FLOPs). In addition, we varied the depth and width of the network to build a compact variant network. It also outperforms other compact variants.

KW - DenseNet

KW - SENet

KW - deep learning

KW - group convolution

KW - keyword spotting

UR - http://www.scopus.com/inward/record.url?scp=85124597644&partnerID=8YFLogxK

U2 - 10.1109/ICECS53924.2021.9665618

DO - 10.1109/ICECS53924.2021.9665618

M3 - 會議論文篇章

AN - SCOPUS:85124597644

T3 - 2021 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Proceedings

BT - 2021 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021

Y2 - 28 November 2021 through 1 December 2021

ER -

Tsai TH, Lin XH. Reduced Model Size Deep Convolutional Neural Networks for Small-Footprint Keyword Spotting. In 2021 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Proceedings. Institute of Electrical and Electronics Engineers Inc. 2021. (2021 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Proceedings). doi: 10.1109/ICECS53924.2021.9665618

Reduced Model Size Deep Convolutional Neural Networks for Small-Footprint Keyword Spotting

Abstract

Publication series

Conference

Keywords

Access to Document

Fingerprint

Cite this