Third-order moments of filtered speech signals for robust speech recognition

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

Third-order moments of filtered speech signals for robust speech recognition

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

Electrical and Computer Engineering

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

2 Scopus citations

Abstract

Novel speech features calculated from third-order statistics of subband-filtered speech signals are introduced and studied for robust speech recognition. These features have the potential to capture nonlinear information not represented by cepstral coefficients. Also, because the features presented in this paper are based on the third-order moments, they may be more immune to Gaussian noise than cepstrals, as Gaussian distributions have zero third-order moments. Experiments on the AURORA2 database studying these features in combination with Mel-frequency cepstral coefficients (MFCC's) are presented, and some improvement over the MFCC-only baseline is shown when clean speech is used for training, though the same improvement is not seen when multi-condition training data is used.

Original language	English
Title of host publication	Nonlinear Analyses and Algorithms for Speech Processing - International Conference on Non-Linear Speech Processing, NOLISP 2005, Revised Selected Papers
Pages	277-283
Number of pages	7
State	Published - 2005
Event	International Conference on Non-Linear Speech Processing, NOLISP 2005 - Barcelona, Spain Duration: Apr 19 2005 → Apr 22 2005

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	3817 LNAI
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	International Conference on Non-Linear Speech Processing, NOLISP 2005
Country/Territory	Spain
City	Barcelona
Period	4/19/05 → 4/22/05

Bibliographical note

Funding Information:
The material in this project was prepared under Grant No. SS-41-83-07 from the Social Science Research Council. Researchers undertaking such projects under Council sponsorship are encouraged to express freely their professional judgment. Therefore, the points of view or opinions stated herein do not necessarily represent the official position or policy of the Social Science Research Council.

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Cite this

Indrebo, K. M., Povinelli, R. J., & Johnson, M. T. (2005). Third-order moments of filtered speech signals for robust speech recognition. In Nonlinear Analyses and Algorithms for Speech Processing - International Conference on Non-Linear Speech Processing, NOLISP 2005, Revised Selected Papers (pp. 277-283). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 3817 LNAI).

Indrebo, Kevin M. ; Povinelli, Richard J. ; Johnson, Michael T. / Third-order moments of filtered speech signals for robust speech recognition. Nonlinear Analyses and Algorithms for Speech Processing - International Conference on Non-Linear Speech Processing, NOLISP 2005, Revised Selected Papers. 2005. pp. 277-283 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{57d21af54bab42db80bc74dc9679a375,

title = "Third-order moments of filtered speech signals for robust speech recognition",

abstract = "Novel speech features calculated from third-order statistics of subband-filtered speech signals are introduced and studied for robust speech recognition. These features have the potential to capture nonlinear information not represented by cepstral coefficients. Also, because the features presented in this paper are based on the third-order moments, they may be more immune to Gaussian noise than cepstrals, as Gaussian distributions have zero third-order moments. Experiments on the AURORA2 database studying these features in combination with Mel-frequency cepstral coefficients (MFCC's) are presented, and some improvement over the MFCC-only baseline is shown when clean speech is used for training, though the same improvement is not seen when multi-condition training data is used.",

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

note = "Funding Information: The material in this project was prepared under Grant No. SS-41-83-07 from the Social Science Research Council. Researchers undertaking such projects under Council sponsorship are encouraged to express freely their professional judgment. Therefore, the points of view or opinions stated herein do not necessarily represent the official position or policy of the Social Science Research Council.; International Conference on Non-Linear Speech Processing, NOLISP 2005 ; Conference date: 19-04-2005 Through 22-04-2005",

year = "2005",

language = "English",

isbn = "3540312579",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

pages = "277--283",

booktitle = "Nonlinear Analyses and Algorithms for Speech Processing - International Conference on Non-Linear Speech Processing, NOLISP 2005, Revised Selected Papers",

}

Indrebo, KM, Povinelli, RJ & Johnson, MT 2005, Third-order moments of filtered speech signals for robust speech recognition. in Nonlinear Analyses and Algorithms for Speech Processing - International Conference on Non-Linear Speech Processing, NOLISP 2005, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 3817 LNAI, pp. 277-283, International Conference on Non-Linear Speech Processing, NOLISP 2005, Barcelona, Spain, 4/19/05.

Third-order moments of filtered speech signals for robust speech recognition. / Indrebo, Kevin M.; Povinelli, Richard J.; Johnson, Michael T.
Nonlinear Analyses and Algorithms for Speech Processing - International Conference on Non-Linear Speech Processing, NOLISP 2005, Revised Selected Papers. 2005. p. 277-283 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 3817 LNAI).

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

TY - GEN

T1 - Third-order moments of filtered speech signals for robust speech recognition

AU - Indrebo, Kevin M.

AU - Povinelli, Richard J.

AU - Johnson, Michael T.

N1 - Funding Information: The material in this project was prepared under Grant No. SS-41-83-07 from the Social Science Research Council. Researchers undertaking such projects under Council sponsorship are encouraged to express freely their professional judgment. Therefore, the points of view or opinions stated herein do not necessarily represent the official position or policy of the Social Science Research Council.

PY - 2005

Y1 - 2005

N2 - Novel speech features calculated from third-order statistics of subband-filtered speech signals are introduced and studied for robust speech recognition. These features have the potential to capture nonlinear information not represented by cepstral coefficients. Also, because the features presented in this paper are based on the third-order moments, they may be more immune to Gaussian noise than cepstrals, as Gaussian distributions have zero third-order moments. Experiments on the AURORA2 database studying these features in combination with Mel-frequency cepstral coefficients (MFCC's) are presented, and some improvement over the MFCC-only baseline is shown when clean speech is used for training, though the same improvement is not seen when multi-condition training data is used.

AB - Novel speech features calculated from third-order statistics of subband-filtered speech signals are introduced and studied for robust speech recognition. These features have the potential to capture nonlinear information not represented by cepstral coefficients. Also, because the features presented in this paper are based on the third-order moments, they may be more immune to Gaussian noise than cepstrals, as Gaussian distributions have zero third-order moments. Experiments on the AURORA2 database studying these features in combination with Mel-frequency cepstral coefficients (MFCC's) are presented, and some improvement over the MFCC-only baseline is shown when clean speech is used for training, though the same improvement is not seen when multi-condition training data is used.

UR - http://www.scopus.com/inward/record.url?scp=33745478447&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33745478447&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:33745478447

SN - 3540312579

SN - 9783540312574

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 277

EP - 283

BT - Nonlinear Analyses and Algorithms for Speech Processing - International Conference on Non-Linear Speech Processing, NOLISP 2005, Revised Selected Papers

T2 - International Conference on Non-Linear Speech Processing, NOLISP 2005

Y2 - 19 April 2005 through 22 April 2005

ER -

Indrebo KM, Povinelli RJ, Johnson MT. Third-order moments of filtered speech signals for robust speech recognition. In Nonlinear Analyses and Algorithms for Speech Processing - International Conference on Non-Linear Speech Processing, NOLISP 2005, Revised Selected Papers. 2005. p. 277-283. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

Third-order moments of filtered speech signals for robust speech recognition

Abstract

Publication series

Conference

Bibliographical note

ASJC Scopus subject areas

Other files and links

Fingerprint

Cite this