WHAT DO NEURAL NETWORKS LISTEN TO? EXPLORING THE CRUCIAL BANDS IN SPEECH ENHANCEMENT USING SINC-CONVOLUTION

Kuan Hsun Ho; Jeih Weih Hung; Berlin Chen

doi:10.1109/ICASSP48485.2024.10445878

WHAT DO NEURAL NETWORKS LISTEN TO? EXPLORING THE CRUCIAL BANDS IN SPEECH ENHANCEMENT USING SINC-CONVOLUTION

Kuan Hsun Ho^*, Jeih Weih Hung, Berlin Chen

^*Corresponding author for this work

Department of Computer Science and Information Engineering

Research output: Contribution to journal › Conference article › peer-review

1 Citation (Scopus)

Abstract

This study introduces a reformed Sinc-convolution (Sincconv) framework tailored for the encoder component of deep networks for speech enhancement (SE). The reformed Sincconv, based on parametrized sinc functions as band-pass filters, offers notable advantages in terms of training efficiency, filter diversity, and interpretability. The reformed Sinc-conv is evaluated in conjunction with various SE models, showcasing its ability to boost SE performance. Furthermore, the reformed Sincconv provides valuable insights into the specific frequency components that are prioritized in an SE scenario. This opens up a new direction of SE research and improving our knowledge of their operating dynamics.

Original language	English
Pages (from-to)	10406-10410
Number of pages	5
Journal	ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
DOIs	https://doi.org/10.1109/ICASSP48485.2024.10445878
Publication status	Published - 2024
Event	2024 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2024 - Seoul, Korea, Republic of Duration: 2024 Apr 14 → 2024 Apr 19

Keywords

Interpretability
Sinc-convolution
Speech Ehancement

ASJC Scopus subject areas

Software
Signal Processing
Electrical and Electronic Engineering

Access to Document

10.1109/ICASSP48485.2024.10445878

Cite this

@article{dd7236e4427c44c7ad464b103eb2579a,

title = "WHAT DO NEURAL NETWORKS LISTEN TO? EXPLORING THE CRUCIAL BANDS IN SPEECH ENHANCEMENT USING SINC-CONVOLUTION",

abstract = "This study introduces a reformed Sinc-convolution (Sincconv) framework tailored for the encoder component of deep networks for speech enhancement (SE). The reformed Sincconv, based on parametrized sinc functions as band-pass filters, offers notable advantages in terms of training efficiency, filter diversity, and interpretability. The reformed Sinc-conv is evaluated in conjunction with various SE models, showcasing its ability to boost SE performance. Furthermore, the reformed Sincconv provides valuable insights into the specific frequency components that are prioritized in an SE scenario. This opens up a new direction of SE research and improving our knowledge of their operating dynamics.",

keywords = "Interpretability, Sinc-convolution, Speech Ehancement",

author = "Ho, {Kuan Hsun} and Hung, {Jeih Weih} and Berlin Chen",

note = "Publisher Copyright: {\textcopyright}2024 IEEE.; 2024 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2024 ; Conference date: 14-04-2024 Through 19-04-2024",

year = "2024",

doi = "10.1109/ICASSP48485.2024.10445878",

language = "English",

pages = "10406--10410",

journal = "ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",

issn = "1520-6149",

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

}

TY - JOUR

T1 - WHAT DO NEURAL NETWORKS LISTEN TO? EXPLORING THE CRUCIAL BANDS IN SPEECH ENHANCEMENT USING SINC-CONVOLUTION

AU - Ho, Kuan Hsun

AU - Hung, Jeih Weih

AU - Chen, Berlin

PY - 2024

Y1 - 2024

N2 - This study introduces a reformed Sinc-convolution (Sincconv) framework tailored for the encoder component of deep networks for speech enhancement (SE). The reformed Sincconv, based on parametrized sinc functions as band-pass filters, offers notable advantages in terms of training efficiency, filter diversity, and interpretability. The reformed Sinc-conv is evaluated in conjunction with various SE models, showcasing its ability to boost SE performance. Furthermore, the reformed Sincconv provides valuable insights into the specific frequency components that are prioritized in an SE scenario. This opens up a new direction of SE research and improving our knowledge of their operating dynamics.

AB - This study introduces a reformed Sinc-convolution (Sincconv) framework tailored for the encoder component of deep networks for speech enhancement (SE). The reformed Sincconv, based on parametrized sinc functions as band-pass filters, offers notable advantages in terms of training efficiency, filter diversity, and interpretability. The reformed Sinc-conv is evaluated in conjunction with various SE models, showcasing its ability to boost SE performance. Furthermore, the reformed Sincconv provides valuable insights into the specific frequency components that are prioritized in an SE scenario. This opens up a new direction of SE research and improving our knowledge of their operating dynamics.

KW - Interpretability

KW - Sinc-convolution

KW - Speech Ehancement

UR - http://www.scopus.com/inward/record.url?scp=105001480921&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=105001480921&partnerID=8YFLogxK

U2 - 10.1109/ICASSP48485.2024.10445878

DO - 10.1109/ICASSP48485.2024.10445878

M3 - Conference article

AN - SCOPUS:105001480921

SN - 1520-6149

SP - 10406

EP - 10410

JO - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

JF - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

T2 - 2024 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2024

Y2 - 14 April 2024 through 19 April 2024

ER -

WHAT DO NEURAL NETWORKS LISTEN TO? EXPLORING THE CRUCIAL BANDS IN SPEECH ENHANCEMENT USING SINC-CONVOLUTION

Abstract

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this