An Exploration of Task-Decoupling on Two-Stage Neural Post Filter for Real-Time Personalized Acoustic Echo Cancellation

Zihan Zhang; Jiayao Sun; Xianjun Xia; Ziqian Wang; Xiaopeng Yan; Yijian Xiao; Lei Xie

doi:10.1109/ASRU57964.2023.10389770

An Exploration of Task-Decoupling on Two-Stage Neural Post Filter for Real-Time Personalized Acoustic Echo Cancellation

Zihan Zhang
, Jiayao Sun
, Xianjun Xia
, Ziqian Wang
, Xiaopeng Yan
, Yijian Xiao
, Lei Xie

School of Computer Science

Northwestern Polytechnical University Xian
ByteDance Ltd.

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

1 Scopus citations

Abstract

Deep learning based techniques have been popularly adopted in acoustic echo cancellation (AEC). Utilization of speaker representation has extended the frontier of AEC, thus attracting many researchers' interest in personalized acoustic echo cancellation (PAEC). Meanwhile, task-decoupling strategies are widely adopted in speech enhancement. To further explore the task-decoupling approach, we propose to use a two-stage task-decoupling post-filter (TDPF) in PAEC. Furthermore, a multi-scale local-global speaker representation is applied to improve speaker extraction in PAEC. Experimental results indicate that the task-decoupling model can yield better performance than a single joint network. The optimal approach is to decouple the echo cancellation from noise and interference speech suppression. Based on the task-decoupling sequence, optimal training strategies for the two-stage model are explored afterwards.

Original language	English
Title of host publication	2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9798350306897
DOIs	https://doi.org/10.1109/ASRU57964.2023.10389770
State	Published - 2023
Event	2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023 - Taipei, Taiwan, Province of China Duration: 16 Dec 2023 → 20 Dec 2023

Publication series

Name	2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023

Conference

Conference	2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023
Country/Territory	Taiwan, Province of China
City	Taipei
Period	16/12/23 → 20/12/23

Keywords

personalized acoustic echo cancellation
speaker representation
task-decoupling
two-stage

Access to Document

10.1109/ASRU57964.2023.10389770

Cite this

Zhang, Z., Sun, J., Xia, X., Wang, Z., Yan, X., Xiao, Y., & Xie, L. (2023). An Exploration of Task-Decoupling on Two-Stage Neural Post Filter for Real-Time Personalized Acoustic Echo Cancellation. In 2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023 (2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ASRU57964.2023.10389770

Zhang, Zihan ; Sun, Jiayao ; Xia, Xianjun et al. / An Exploration of Task-Decoupling on Two-Stage Neural Post Filter for Real-Time Personalized Acoustic Echo Cancellation. 2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023. Institute of Electrical and Electronics Engineers Inc., 2023. (2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023).

@inproceedings{b05a12d49ca0498fa2833a362a2138f1,

title = "An Exploration of Task-Decoupling on Two-Stage Neural Post Filter for Real-Time Personalized Acoustic Echo Cancellation",

abstract = "Deep learning based techniques have been popularly adopted in acoustic echo cancellation (AEC). Utilization of speaker representation has extended the frontier of AEC, thus attracting many researchers' interest in personalized acoustic echo cancellation (PAEC). Meanwhile, task-decoupling strategies are widely adopted in speech enhancement. To further explore the task-decoupling approach, we propose to use a two-stage task-decoupling post-filter (TDPF) in PAEC. Furthermore, a multi-scale local-global speaker representation is applied to improve speaker extraction in PAEC. Experimental results indicate that the task-decoupling model can yield better performance than a single joint network. The optimal approach is to decouple the echo cancellation from noise and interference speech suppression. Based on the task-decoupling sequence, optimal training strategies for the two-stage model are explored afterwards.",

keywords = "personalized acoustic echo cancellation, speaker representation, task-decoupling, two-stage",

author = "Zihan Zhang and Jiayao Sun and Xianjun Xia and Ziqian Wang and Xiaopeng Yan and Yijian Xiao and Lei Xie",

note = "Publisher Copyright: {\textcopyright} 2023 IEEE.; 2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023 ; Conference date: 16-12-2023 Through 20-12-2023",

year = "2023",

doi = "10.1109/ASRU57964.2023.10389770",

language = "英语",

series = "2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023",

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

booktitle = "2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023",

}

Zhang, Z, Sun, J, Xia, X, Wang, Z, Yan, X, Xiao, Y & Xie, L 2023, An Exploration of Task-Decoupling on Two-Stage Neural Post Filter for Real-Time Personalized Acoustic Echo Cancellation. in 2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023. 2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023, Institute of Electrical and Electronics Engineers Inc., 2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023, Taipei, Taiwan, Province of China, 16/12/23. https://doi.org/10.1109/ASRU57964.2023.10389770

An Exploration of Task-Decoupling on Two-Stage Neural Post Filter for Real-Time Personalized Acoustic Echo Cancellation. / Zhang, Zihan; Sun, Jiayao; Xia, Xianjun et al.
2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023. Institute of Electrical and Electronics Engineers Inc., 2023. (2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023).

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

TY - GEN

T1 - An Exploration of Task-Decoupling on Two-Stage Neural Post Filter for Real-Time Personalized Acoustic Echo Cancellation

AU - Zhang, Zihan

AU - Sun, Jiayao

AU - Xia, Xianjun

AU - Wang, Ziqian

AU - Yan, Xiaopeng

AU - Xiao, Yijian

AU - Xie, Lei

PY - 2023

Y1 - 2023

N2 - Deep learning based techniques have been popularly adopted in acoustic echo cancellation (AEC). Utilization of speaker representation has extended the frontier of AEC, thus attracting many researchers' interest in personalized acoustic echo cancellation (PAEC). Meanwhile, task-decoupling strategies are widely adopted in speech enhancement. To further explore the task-decoupling approach, we propose to use a two-stage task-decoupling post-filter (TDPF) in PAEC. Furthermore, a multi-scale local-global speaker representation is applied to improve speaker extraction in PAEC. Experimental results indicate that the task-decoupling model can yield better performance than a single joint network. The optimal approach is to decouple the echo cancellation from noise and interference speech suppression. Based on the task-decoupling sequence, optimal training strategies for the two-stage model are explored afterwards.

AB - Deep learning based techniques have been popularly adopted in acoustic echo cancellation (AEC). Utilization of speaker representation has extended the frontier of AEC, thus attracting many researchers' interest in personalized acoustic echo cancellation (PAEC). Meanwhile, task-decoupling strategies are widely adopted in speech enhancement. To further explore the task-decoupling approach, we propose to use a two-stage task-decoupling post-filter (TDPF) in PAEC. Furthermore, a multi-scale local-global speaker representation is applied to improve speaker extraction in PAEC. Experimental results indicate that the task-decoupling model can yield better performance than a single joint network. The optimal approach is to decouple the echo cancellation from noise and interference speech suppression. Based on the task-decoupling sequence, optimal training strategies for the two-stage model are explored afterwards.

KW - personalized acoustic echo cancellation

KW - speaker representation

KW - task-decoupling

KW - two-stage

UR - https://www.scopus.com/pages/publications/85184657027

U2 - 10.1109/ASRU57964.2023.10389770

DO - 10.1109/ASRU57964.2023.10389770

M3 - 会议稿件

AN - SCOPUS:85184657027

T3 - 2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023

BT - 2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023

Y2 - 16 December 2023 through 20 December 2023

ER -

Zhang Z, Sun J, Xia X, Wang Z, Yan X, Xiao Y et al. An Exploration of Task-Decoupling on Two-Stage Neural Post Filter for Real-Time Personalized Acoustic Echo Cancellation. In 2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023. Institute of Electrical and Electronics Engineers Inc. 2023. (2023 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023). doi: 10.1109/ASRU57964.2023.10389770

An Exploration of Task-Decoupling on Two-Stage Neural Post Filter for Real-Time Personalized Acoustic Echo Cancellation

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this