A Joint Approach for Single-Channel Speaker Identification and Speech Separation

Pejman Mowlaee, Rahim Saeidi, Mads Græsbøll Christensen, Zheng-Hua Tan, Tomi Kinnunen, Pasi Franti, Søren Holdt Jensen

Research output: Contribution to journalJournal articleResearchpeer-review

27 Citations (Scopus)
465 Downloads (Pure)

Abstract

In this paper, we present a novel system for joint speaker identification and speech separation. For speaker identification a single-channel speaker identification algorithm is proposed which provides an estimate of signal-to-signal ratio (SSR) as a by-product. For speech separation, we propose a sinusoidal model-based algorithm. The speech separation algorithm consists of a double-talk/single-talk detector followed by a minimum mean square error estimator of sinusoidal parameters for finding optimal codevectors from pre-trained speaker codebooks. In evaluating the proposed system, we start from a situation where we have prior information of codebook indices, speaker identities and SSR-level, and then, by relaxing these assumptions one by one, we demonstrate the efficiency of the proposed fully blind system. In contrast to previous studies that mostly focus on automatic speech recognition (ASR) accuracy, here, we report the objective and subjective results as well. The results show that the proposed system performs as well as the best of the state-of-the-art in terms of perceived quality while its performance in terms of speaker identification and automatic speech recognition results are generally lower. It outperforms the state-of-the-art in terms of intelligibility showing that the ASR results are not conclusive. The proposed method achieves on average, 52.3% ASR accuracy, 41.2 points in MUSHRA and 85.9% in speech intelligibility.
Original languageEnglish
JournalI E E E Transactions on Audio, Speech and Language Processing
Volume20
Issue number9
Pages (from-to)2586 - 2601
Number of pages16
ISSN1558-7916
DOIs
Publication statusPublished - Nov 2012

Fingerprint

speech recognition
Speech recognition
Speech intelligibility
intelligibility
estimators
Mean square error
Byproducts
Detectors
detectors
estimates

Keywords

  • BSS EVAL
  • single-channel speech separation
  • sinusoidal modeling
  • speaker identificatio
  • speech recognition

Cite this

@article{5a431878158e47408c82b0c87ac87d53,
title = "A Joint Approach for Single-Channel Speaker Identification and Speech Separation",
abstract = "In this paper, we present a novel system for joint speaker identification and speech separation. For speaker identification a single-channel speaker identification algorithm is proposed which provides an estimate of signal-to-signal ratio (SSR) as a by-product. For speech separation, we propose a sinusoidal model-based algorithm. The speech separation algorithm consists of a double-talk/single-talk detector followed by a minimum mean square error estimator of sinusoidal parameters for finding optimal codevectors from pre-trained speaker codebooks. In evaluating the proposed system, we start from a situation where we have prior information of codebook indices, speaker identities and SSR-level, and then, by relaxing these assumptions one by one, we demonstrate the efficiency of the proposed fully blind system. In contrast to previous studies that mostly focus on automatic speech recognition (ASR) accuracy, here, we report the objective and subjective results as well. The results show that the proposed system performs as well as the best of the state-of-the-art in terms of perceived quality while its performance in terms of speaker identification and automatic speech recognition results are generally lower. It outperforms the state-of-the-art in terms of intelligibility showing that the ASR results are not conclusive. The proposed method achieves on average, 52.3{\%} ASR accuracy, 41.2 points in MUSHRA and 85.9{\%} in speech intelligibility.",
keywords = "BSS EVAL, single-channel speech separation, sinusoidal modeling, speaker identificatio, speech recognition",
author = "Pejman Mowlaee and Rahim Saeidi and Christensen, {Mads Gr{\ae}sb{\o}ll} and Zheng-Hua Tan and Tomi Kinnunen and Pasi Franti and Jensen, {S{\o}ren Holdt}",
year = "2012",
month = "11",
doi = "10.1109/TASL.2012.2208627",
language = "English",
volume = "20",
pages = "2586 -- 2601",
journal = "IEEE/ACM Transactions on Audio, Speech, and Language Processing",
issn = "2329-9290",
publisher = "IEEE Signal Processing Society",
number = "9",

}

A Joint Approach for Single-Channel Speaker Identification and Speech Separation. / Mowlaee, Pejman; Saeidi, Rahim; Christensen, Mads Græsbøll; Tan, Zheng-Hua; Kinnunen, Tomi; Franti, Pasi; Jensen, Søren Holdt.

In: I E E E Transactions on Audio, Speech and Language Processing, Vol. 20, No. 9, 11.2012, p. 2586 - 2601.

Research output: Contribution to journalJournal articleResearchpeer-review

TY - JOUR

T1 - A Joint Approach for Single-Channel Speaker Identification and Speech Separation

AU - Mowlaee, Pejman

AU - Saeidi, Rahim

AU - Christensen, Mads Græsbøll

AU - Tan, Zheng-Hua

AU - Kinnunen, Tomi

AU - Franti, Pasi

AU - Jensen, Søren Holdt

PY - 2012/11

Y1 - 2012/11

N2 - In this paper, we present a novel system for joint speaker identification and speech separation. For speaker identification a single-channel speaker identification algorithm is proposed which provides an estimate of signal-to-signal ratio (SSR) as a by-product. For speech separation, we propose a sinusoidal model-based algorithm. The speech separation algorithm consists of a double-talk/single-talk detector followed by a minimum mean square error estimator of sinusoidal parameters for finding optimal codevectors from pre-trained speaker codebooks. In evaluating the proposed system, we start from a situation where we have prior information of codebook indices, speaker identities and SSR-level, and then, by relaxing these assumptions one by one, we demonstrate the efficiency of the proposed fully blind system. In contrast to previous studies that mostly focus on automatic speech recognition (ASR) accuracy, here, we report the objective and subjective results as well. The results show that the proposed system performs as well as the best of the state-of-the-art in terms of perceived quality while its performance in terms of speaker identification and automatic speech recognition results are generally lower. It outperforms the state-of-the-art in terms of intelligibility showing that the ASR results are not conclusive. The proposed method achieves on average, 52.3% ASR accuracy, 41.2 points in MUSHRA and 85.9% in speech intelligibility.

AB - In this paper, we present a novel system for joint speaker identification and speech separation. For speaker identification a single-channel speaker identification algorithm is proposed which provides an estimate of signal-to-signal ratio (SSR) as a by-product. For speech separation, we propose a sinusoidal model-based algorithm. The speech separation algorithm consists of a double-talk/single-talk detector followed by a minimum mean square error estimator of sinusoidal parameters for finding optimal codevectors from pre-trained speaker codebooks. In evaluating the proposed system, we start from a situation where we have prior information of codebook indices, speaker identities and SSR-level, and then, by relaxing these assumptions one by one, we demonstrate the efficiency of the proposed fully blind system. In contrast to previous studies that mostly focus on automatic speech recognition (ASR) accuracy, here, we report the objective and subjective results as well. The results show that the proposed system performs as well as the best of the state-of-the-art in terms of perceived quality while its performance in terms of speaker identification and automatic speech recognition results are generally lower. It outperforms the state-of-the-art in terms of intelligibility showing that the ASR results are not conclusive. The proposed method achieves on average, 52.3% ASR accuracy, 41.2 points in MUSHRA and 85.9% in speech intelligibility.

KW - BSS EVAL

KW - single-channel speech separation

KW - sinusoidal modeling

KW - speaker identificatio

KW - speech recognition

UR - http://www.scopus.com/inward/record.url?scp=84865693405&partnerID=8YFLogxK

U2 - 10.1109/TASL.2012.2208627

DO - 10.1109/TASL.2012.2208627

M3 - Journal article

VL - 20

SP - 2586

EP - 2601

JO - IEEE/ACM Transactions on Audio, Speech, and Language Processing

JF - IEEE/ACM Transactions on Audio, Speech, and Language Processing

SN - 2329-9290

IS - 9

ER -