Polynomial eigenvalue decomposition-based target speaker voice activity detection in the presence of competing talkers

Vincent W. Neo; Stephan Weiss; Simon W. McKnight; Aidan O. T. Hogg; Patrick A. Naylor

doi:10.1109/IWAENC53105.2022.9914796

Polynomial eigenvalue decomposition-based target speaker voice activity detection in the presence of competing talkers

Vincent W. Neo, Stephan Weiss, Simon W. McKnight, Aidan O. T. Hogg, Patrick A. Naylor

Electronic And Electrical Engineering

Research output: Contribution to conference › Paper › peer-review

4 Citations (Scopus)

7 Downloads (Pure)

Abstract

Voice activity detection (VAD) algorithms are essential for many speech processing applications, such as speaker diarization, automatic speech recognition, speech enhancement, and speech coding. With a good VAD algorithm, non-speech segments can be excluded to improve the performance and computation of these applications. In this paper, we propose a polynomial eigenvalue decomposition-based target-speaker VAD algorithm to detect unseen target speakers in the presence of competing talkers. The proposed approach uses frame-based processing across multi-microphones to compute the syndrome energy, used for testing the presence or absence of a target speaker. The proposed approach is consistently among the best in F1 and balanced accuracy scores over the investigated range of signal to interference ratio (SIR) from -10 dB to 20 dB.

Original language	English
Pages	1-5
Number of pages	5
DOIs	https://doi.org/10.1109/IWAENC53105.2022.9914796
Publication status	Published - 17 Oct 2022
Event	17th International Workshop on Acoustic Signal Enhancement - Bamberg, Germany Duration: 5 Sep 2022 → 8 Sep 2022

Conference

Conference	17th International Workshop on Acoustic Signal Enhancement
Abbreviated title	IWAENC 2022
Country/Territory	Germany
City	Bamberg
Period	5/09/22 → 8/09/22

Keywords

polynomial eigenvalue decomposition
target speaker voice activity detection
speaker activity detection

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1109/IWAENC53105.2022.9914796

Neo-etal-IWAENC2022-Polynomial-eigenvalue-decomposition-based-target-speaker-voice-activity-detectionAccepted author manuscript, 836 KBLicence: Strath_1

Cite this

@conference{4ad77e8f59d0438abe78c1c4db02cb29,

title = "Polynomial eigenvalue decomposition-based target speaker voice activity detection in the presence of competing talkers",

abstract = "Voice activity detection (VAD) algorithms are essential for many speech processing applications, such as speaker diarization, automatic speech recognition, speech enhancement, and speech coding. With a good VAD algorithm, non-speech segments can be excluded to improve the performance and computation of these applications. In this paper, we propose a polynomial eigenvalue decomposition-based target-speaker VAD algorithm to detect unseen target speakers in the presence of competing talkers. The proposed approach uses frame-based processing across multi-microphones to compute the syndrome energy, used for testing the presence or absence of a target speaker. The proposed approach is consistently among the best in F1 and balanced accuracy scores over the investigated range of signal to interference ratio (SIR) from -10 dB to 20 dB.",

keywords = "polynomial eigenvalue decomposition, target speaker voice activity detection, speaker activity detection",

author = "Neo, {Vincent W.} and Stephan Weiss and McKnight, {Simon W.} and Hogg, {Aidan O. T.} and Naylor, {Patrick A.}",

note = "{\textcopyright} 2022 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting /republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.; 17th International Workshop on Acoustic Signal Enhancement, IWAENC 2022 ; Conference date: 05-09-2022 Through 08-09-2022",

year = "2022",

month = oct,

day = "17",

doi = "10.1109/IWAENC53105.2022.9914796",

language = "English",

pages = "1--5",

}

TY - CONF

T1 - Polynomial eigenvalue decomposition-based target speaker voice activity detection in the presence of competing talkers

AU - Neo, Vincent W.

AU - Weiss, Stephan

AU - McKnight, Simon W.

AU - Hogg, Aidan O. T.

AU - Naylor, Patrick A.

N1 - © 2022 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting /republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

PY - 2022/10/17

Y1 - 2022/10/17

N2 - Voice activity detection (VAD) algorithms are essential for many speech processing applications, such as speaker diarization, automatic speech recognition, speech enhancement, and speech coding. With a good VAD algorithm, non-speech segments can be excluded to improve the performance and computation of these applications. In this paper, we propose a polynomial eigenvalue decomposition-based target-speaker VAD algorithm to detect unseen target speakers in the presence of competing talkers. The proposed approach uses frame-based processing across multi-microphones to compute the syndrome energy, used for testing the presence or absence of a target speaker. The proposed approach is consistently among the best in F1 and balanced accuracy scores over the investigated range of signal to interference ratio (SIR) from -10 dB to 20 dB.

AB - Voice activity detection (VAD) algorithms are essential for many speech processing applications, such as speaker diarization, automatic speech recognition, speech enhancement, and speech coding. With a good VAD algorithm, non-speech segments can be excluded to improve the performance and computation of these applications. In this paper, we propose a polynomial eigenvalue decomposition-based target-speaker VAD algorithm to detect unseen target speakers in the presence of competing talkers. The proposed approach uses frame-based processing across multi-microphones to compute the syndrome energy, used for testing the presence or absence of a target speaker. The proposed approach is consistently among the best in F1 and balanced accuracy scores over the investigated range of signal to interference ratio (SIR) from -10 dB to 20 dB.

KW - polynomial eigenvalue decomposition

KW - target speaker voice activity detection

KW - speaker activity detection

UR - https://iwaenc2022.org/

U2 - 10.1109/IWAENC53105.2022.9914796

DO - 10.1109/IWAENC53105.2022.9914796

M3 - Paper

SP - 1

EP - 5

T2 - 17th International Workshop on Acoustic Signal Enhancement

Y2 - 5 September 2022 through 8 September 2022

ER -

Polynomial eigenvalue decomposition-based target speaker voice activity detection in the presence of competing talkers

Abstract

Conference

Keywords

UN SDGs

Access to Document

Other files and links

Fingerprint

Signal Processing in the Information Age (UDRC III)

Cite this

Polynomial eigenvalue decomposition-based target speaker voice activity detection in the presence of competing talkers

Abstract

Conference

Keywords

UN SDGs

Access to Document

Other files and links

Fingerprint

Projects

Signal Processing in the Information Age (UDRC III)

Cite this