Impact of phase estimation on single-channel speech separation based on time-frequency masking

Florian Mayer, Donald S. Williamson, Pejman Mowlaee, Deliang Wang

Research output: Contribution to journalArticleResearchpeer-review

Abstract

Time-frequency masking is a common solution for the single-channel source separation (SCSS) problem where the goal is to find a time-frequency mask that separates the underlying sources from an observed mixture. An estimated mask is then applied to the mixed signal to extract the desired signal. During signal reconstruction, the time-frequency-masked spectral amplitude is combined with the mixture phase. This article considers the impact of replacing the mixture spectral phase with an estimated clean spectral phase combined with the estimated magnitude spectrum using a conventional model-based approach. As the proposed phase estimator requires estimated fundamental frequency of the underlying signal from the mixture, a robust pitch estimator is proposed. The upper-bound clean phase results show the potential of phase-Aware processing in single-channel source separation. Also, the experiments demonstrate that replacing the mixture phase with the estimated clean spectral phase consistently improves perceptual speech quality, predicted speech intelligibility, and source separation performance across all signal-To-noise ratio and noise scenarios.

Original languageEnglish
Pages (from-to)4668-4679
Number of pages12
JournalJournal of the Acoustical Society of America
Volume141
Issue number6
DOIs
Publication statusPublished - 1 Jun 2017

Fingerprint

masking
estimators
masks
intelligibility
signal to noise ratios
Masking
Spectrality

ASJC Scopus subject areas

  • Arts and Humanities (miscellaneous)
  • Acoustics and Ultrasonics

Cite this

Impact of phase estimation on single-channel speech separation based on time-frequency masking. / Mayer, Florian; Williamson, Donald S.; Mowlaee, Pejman; Wang, Deliang.

In: Journal of the Acoustical Society of America, Vol. 141, No. 6, 01.06.2017, p. 4668-4679.

Research output: Contribution to journalArticleResearchpeer-review

Mayer, Florian ; Williamson, Donald S. ; Mowlaee, Pejman ; Wang, Deliang. / Impact of phase estimation on single-channel speech separation based on time-frequency masking. In: Journal of the Acoustical Society of America. 2017 ; Vol. 141, No. 6. pp. 4668-4679.
@article{c365c8933d1e4ad8b98a2a492ceafaf6,
title = "Impact of phase estimation on single-channel speech separation based on time-frequency masking",
abstract = "Time-frequency masking is a common solution for the single-channel source separation (SCSS) problem where the goal is to find a time-frequency mask that separates the underlying sources from an observed mixture. An estimated mask is then applied to the mixed signal to extract the desired signal. During signal reconstruction, the time-frequency-masked spectral amplitude is combined with the mixture phase. This article considers the impact of replacing the mixture spectral phase with an estimated clean spectral phase combined with the estimated magnitude spectrum using a conventional model-based approach. As the proposed phase estimator requires estimated fundamental frequency of the underlying signal from the mixture, a robust pitch estimator is proposed. The upper-bound clean phase results show the potential of phase-Aware processing in single-channel source separation. Also, the experiments demonstrate that replacing the mixture phase with the estimated clean spectral phase consistently improves perceptual speech quality, predicted speech intelligibility, and source separation performance across all signal-To-noise ratio and noise scenarios.",
author = "Florian Mayer and Williamson, {Donald S.} and Pejman Mowlaee and Deliang Wang",
year = "2017",
month = "6",
day = "1",
doi = "10.1121/1.4986647",
language = "English",
volume = "141",
pages = "4668--4679",
journal = "The journal of the Acoustical Society of America",
issn = "0001-4966",
publisher = "Acoustical Society of America",
number = "6",

}

TY - JOUR

T1 - Impact of phase estimation on single-channel speech separation based on time-frequency masking

AU - Mayer, Florian

AU - Williamson, Donald S.

AU - Mowlaee, Pejman

AU - Wang, Deliang

PY - 2017/6/1

Y1 - 2017/6/1

N2 - Time-frequency masking is a common solution for the single-channel source separation (SCSS) problem where the goal is to find a time-frequency mask that separates the underlying sources from an observed mixture. An estimated mask is then applied to the mixed signal to extract the desired signal. During signal reconstruction, the time-frequency-masked spectral amplitude is combined with the mixture phase. This article considers the impact of replacing the mixture spectral phase with an estimated clean spectral phase combined with the estimated magnitude spectrum using a conventional model-based approach. As the proposed phase estimator requires estimated fundamental frequency of the underlying signal from the mixture, a robust pitch estimator is proposed. The upper-bound clean phase results show the potential of phase-Aware processing in single-channel source separation. Also, the experiments demonstrate that replacing the mixture phase with the estimated clean spectral phase consistently improves perceptual speech quality, predicted speech intelligibility, and source separation performance across all signal-To-noise ratio and noise scenarios.

AB - Time-frequency masking is a common solution for the single-channel source separation (SCSS) problem where the goal is to find a time-frequency mask that separates the underlying sources from an observed mixture. An estimated mask is then applied to the mixed signal to extract the desired signal. During signal reconstruction, the time-frequency-masked spectral amplitude is combined with the mixture phase. This article considers the impact of replacing the mixture spectral phase with an estimated clean spectral phase combined with the estimated magnitude spectrum using a conventional model-based approach. As the proposed phase estimator requires estimated fundamental frequency of the underlying signal from the mixture, a robust pitch estimator is proposed. The upper-bound clean phase results show the potential of phase-Aware processing in single-channel source separation. Also, the experiments demonstrate that replacing the mixture phase with the estimated clean spectral phase consistently improves perceptual speech quality, predicted speech intelligibility, and source separation performance across all signal-To-noise ratio and noise scenarios.

UR - http://www.scopus.com/inward/record.url?scp=85021156150&partnerID=8YFLogxK

U2 - 10.1121/1.4986647

DO - 10.1121/1.4986647

M3 - Article

VL - 141

SP - 4668

EP - 4679

JO - The journal of the Acoustical Society of America

JF - The journal of the Acoustical Society of America

SN - 0001-4966

IS - 6

ER -