Minimum mean-squared error estimation of mel-frequency cepstral coefficients using a novel distortion model

Kevin M. Indrebo; Richard J. Povinelli; Michael T. Johnson

doi:10.1109/TASL.2008.2002083

Minimum mean-squared error estimation of mel-frequency cepstral coefficients using a novel distortion model

Kevin M. Indrebo, Richard J. Povinelli, Michael T. Johnson

Electrical and Computer Engineering

Research output: Contribution to journal › Article › peer-review

21 Scopus citations

Abstract

In this paper, a new method for statistical estimation of Mel-frequency cepstral coefficients (MFCCs) in noisy speech signals is proposed. Previous research has shown that model-based feature domain enhancement of speech signals for use in robust speech recognition can improve recognition accuracy significantly. These methods, which typically work in the log spectral or cepstral domain, must face the high complexity of distortion models caused by the nonlinear interaction of speech and noise in these domains. In this paper, an additive cepstral distortion model (ACDM) is developed, and used with a minimum mean-squared error (MMSE) estimator for recovery of MFCC features corrupted by additive noise. The proposed ACDM-MMSE estimation algorithm is evaluated on the Aurora2 database, and is shown to provide significant improvement in word recognition accuracy over the baseline.

Original language	English
Pages (from-to)	1654-1661
Number of pages	8
Journal	IEEE Transactions on Audio, Speech and Language Processing
Volume	16
Issue number	8
DOIs	https://doi.org/10.1109/TASL.2008.2002083
State	Published - Nov 2008

Bibliographical note

Funding Information:
Manuscript received January 15, 2008; revised May 15, 2008. Current version published October 17, 2008. This work was supported by the Graduate Assistance in Areas of National Need (GAANN) program, funded by the U.S. Department of Education. The associate editor coordinating the review of this manuscript and approving it for publication was Prof. Abeer Alwan.

Keywords

Parameter estimation
Robustness
Speech recognition

ASJC Scopus subject areas

Acoustics and Ultrasonics
Electrical and Electronic Engineering

Access to Document

10.1109/TASL.2008.2002083

Cite this

@article{0357dae557f0441ab0f1d0f1ae72e4d4,

title = "Minimum mean-squared error estimation of mel-frequency cepstral coefficients using a novel distortion model",

abstract = "In this paper, a new method for statistical estimation of Mel-frequency cepstral coefficients (MFCCs) in noisy speech signals is proposed. Previous research has shown that model-based feature domain enhancement of speech signals for use in robust speech recognition can improve recognition accuracy significantly. These methods, which typically work in the log spectral or cepstral domain, must face the high complexity of distortion models caused by the nonlinear interaction of speech and noise in these domains. In this paper, an additive cepstral distortion model (ACDM) is developed, and used with a minimum mean-squared error (MMSE) estimator for recovery of MFCC features corrupted by additive noise. The proposed ACDM-MMSE estimation algorithm is evaluated on the Aurora2 database, and is shown to provide significant improvement in word recognition accuracy over the baseline.",

keywords = "Parameter estimation, Robustness, Speech recognition",

author = "Indrebo, {Kevin M.} and Povinelli, {Richard J.} and Johnson, {Michael T.}",

note = "Funding Information: Manuscript received January 15, 2008; revised May 15, 2008. Current version published October 17, 2008. This work was supported by the Graduate Assistance in Areas of National Need (GAANN) program, funded by the U.S. Department of Education. The associate editor coordinating the review of this manuscript and approving it for publication was Prof. Abeer Alwan.",

year = "2008",

month = nov,

doi = "10.1109/TASL.2008.2002083",

language = "English",

volume = "16",

pages = "1654--1661",

number = "8",

}

TY - JOUR

T1 - Minimum mean-squared error estimation of mel-frequency cepstral coefficients using a novel distortion model

AU - Indrebo, Kevin M.

AU - Povinelli, Richard J.

AU - Johnson, Michael T.

N1 - Funding Information: Manuscript received January 15, 2008; revised May 15, 2008. Current version published October 17, 2008. This work was supported by the Graduate Assistance in Areas of National Need (GAANN) program, funded by the U.S. Department of Education. The associate editor coordinating the review of this manuscript and approving it for publication was Prof. Abeer Alwan.

PY - 2008/11

Y1 - 2008/11

N2 - In this paper, a new method for statistical estimation of Mel-frequency cepstral coefficients (MFCCs) in noisy speech signals is proposed. Previous research has shown that model-based feature domain enhancement of speech signals for use in robust speech recognition can improve recognition accuracy significantly. These methods, which typically work in the log spectral or cepstral domain, must face the high complexity of distortion models caused by the nonlinear interaction of speech and noise in these domains. In this paper, an additive cepstral distortion model (ACDM) is developed, and used with a minimum mean-squared error (MMSE) estimator for recovery of MFCC features corrupted by additive noise. The proposed ACDM-MMSE estimation algorithm is evaluated on the Aurora2 database, and is shown to provide significant improvement in word recognition accuracy over the baseline.

AB - In this paper, a new method for statistical estimation of Mel-frequency cepstral coefficients (MFCCs) in noisy speech signals is proposed. Previous research has shown that model-based feature domain enhancement of speech signals for use in robust speech recognition can improve recognition accuracy significantly. These methods, which typically work in the log spectral or cepstral domain, must face the high complexity of distortion models caused by the nonlinear interaction of speech and noise in these domains. In this paper, an additive cepstral distortion model (ACDM) is developed, and used with a minimum mean-squared error (MMSE) estimator for recovery of MFCC features corrupted by additive noise. The proposed ACDM-MMSE estimation algorithm is evaluated on the Aurora2 database, and is shown to provide significant improvement in word recognition accuracy over the baseline.

KW - Parameter estimation

KW - Robustness

KW - Speech recognition

UR - http://www.scopus.com/inward/record.url?scp=70350461532&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=70350461532&partnerID=8YFLogxK

U2 - 10.1109/TASL.2008.2002083

DO - 10.1109/TASL.2008.2002083

M3 - Article

AN - SCOPUS:70350461532

SN - 1558-7916

VL - 16

SP - 1654

EP - 1661

JO - IEEE Transactions on Audio, Speech and Language Processing

JF - IEEE Transactions on Audio, Speech and Language Processing

IS - 8

ER -

Minimum mean-squared error estimation of mel-frequency cepstral coefficients using a novel distortion model

Abstract

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this