Robust speech feature extraction by growth transformation in reproducing kernel Hilbert space

Shantanu Chakrabartty; Yunbin Deng; Gert Cauwenberghs

Robust speech feature extraction by growth transformation in reproducing kernel Hilbert space

Shantanu Chakrabartty, Yunbin Deng, Gert Cauwenberghs

Research output: Contribution to journal › Conference article › peer-review

Abstract

A robust speech feature extraction procedure, by kernel regression nonlinear predictive coding, is presented. Features maximally insensitive to additive noise are obtained by growth transformation of regression functions spanning a Reproducing Kernel Hubert Space (RKHS). Experiments on TIDIGIT demonstrate consistent robustness of the new features to noise of varying statistics, yielding significant improvements in digit recognition accuracy over identical models trained using Mel-scale cepstral features and evaluated at noise levels between 0 and 30 dB SNR.

Original language	English
Pages (from-to)	I133-I136
Journal	ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Volume	1
State	Published - 2004
Event	Proceedings - IEEE International Conference on Acoustics, Speech, and Signal Processing - Montreal, Que, Canada Duration: May 17 2004 → May 21 2004

Cite this

@article{6ca0c133ff474366b28c1a73352aaff4,

title = "Robust speech feature extraction by growth transformation in reproducing kernel Hilbert space",

abstract = "A robust speech feature extraction procedure, by kernel regression nonlinear predictive coding, is presented. Features maximally insensitive to additive noise are obtained by growth transformation of regression functions spanning a Reproducing Kernel Hubert Space (RKHS). Experiments on TIDIGIT demonstrate consistent robustness of the new features to noise of varying statistics, yielding significant improvements in digit recognition accuracy over identical models trained using Mel-scale cepstral features and evaluated at noise levels between 0 and 30 dB SNR.",

author = "Shantanu Chakrabartty and Yunbin Deng and Gert Cauwenberghs",

year = "2004",

language = "English",

volume = "1",

pages = "I133--I136",

journal = "ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",

issn = "1520-6149",

note = "Proceedings - IEEE International Conference on Acoustics, Speech, and Signal Processing ; Conference date: 17-05-2004 Through 21-05-2004",

}

TY - JOUR

T1 - Robust speech feature extraction by growth transformation in reproducing kernel Hilbert space

AU - Chakrabartty, Shantanu

AU - Deng, Yunbin

AU - Cauwenberghs, Gert

PY - 2004

Y1 - 2004

N2 - A robust speech feature extraction procedure, by kernel regression nonlinear predictive coding, is presented. Features maximally insensitive to additive noise are obtained by growth transformation of regression functions spanning a Reproducing Kernel Hubert Space (RKHS). Experiments on TIDIGIT demonstrate consistent robustness of the new features to noise of varying statistics, yielding significant improvements in digit recognition accuracy over identical models trained using Mel-scale cepstral features and evaluated at noise levels between 0 and 30 dB SNR.

AB - A robust speech feature extraction procedure, by kernel regression nonlinear predictive coding, is presented. Features maximally insensitive to additive noise are obtained by growth transformation of regression functions spanning a Reproducing Kernel Hubert Space (RKHS). Experiments on TIDIGIT demonstrate consistent robustness of the new features to noise of varying statistics, yielding significant improvements in digit recognition accuracy over identical models trained using Mel-scale cepstral features and evaluated at noise levels between 0 and 30 dB SNR.

UR - http://www.scopus.com/inward/record.url?scp=4544318506&partnerID=8YFLogxK

M3 - Conference article

AN - SCOPUS:4544318506

SN - 1520-6149

VL - 1

SP - I133-I136

JO - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

JF - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

T2 - Proceedings - IEEE International Conference on Acoustics, Speech, and Signal Processing

Y2 - 17 May 2004 through 21 May 2004

ER -

Robust speech feature extraction by growth transformation in reproducing kernel Hilbert space

Abstract

Other files and links

Fingerprint

Cite this