Selection of spectral compressive operator for vector Taylor series-based model adaptation in noisy environments

Soonho Baek, Hong Goo Kang

Research output: Contribution to journalArticle

Abstract

This letter investigates the impact of spectral compression on the vector Taylor series-based model adaptation algorithm. Unlike mel-frequency cepstral coefficients obtained by the logarithmic compression, the fractional power compression is used for extracting features. Since the relationship between acoustic models for clean and noisy speech depends on nonlinearity of the spectrum, it is important to select an appropriate compressive operator in the model adaptation. In this letter, the dependency of spectral nonlinearity on the speech recognition system is analyzed in various noisy environments. Experimental results confirm that the replacement of the compressive operator improves the performance of the model adaptation.

Original languageEnglish
Pages (from-to)EL284-EL290
JournalJournal of the Acoustical Society of America
Volume135
Issue number6
DOIs
Publication statusPublished - 2014 Jun

Fingerprint

Taylor series
operators
nonlinearity
speech recognition
acoustics
Operator
Spectrality
coefficients
Compression
Nonlinearity

All Science Journal Classification (ASJC) codes

  • Arts and Humanities (miscellaneous)
  • Acoustics and Ultrasonics

Cite this

@article{d52def2baf45444db7181159caa1b0a7,
title = "Selection of spectral compressive operator for vector Taylor series-based model adaptation in noisy environments",
abstract = "This letter investigates the impact of spectral compression on the vector Taylor series-based model adaptation algorithm. Unlike mel-frequency cepstral coefficients obtained by the logarithmic compression, the fractional power compression is used for extracting features. Since the relationship between acoustic models for clean and noisy speech depends on nonlinearity of the spectrum, it is important to select an appropriate compressive operator in the model adaptation. In this letter, the dependency of spectral nonlinearity on the speech recognition system is analyzed in various noisy environments. Experimental results confirm that the replacement of the compressive operator improves the performance of the model adaptation.",
author = "Soonho Baek and Kang, {Hong Goo}",
year = "2014",
month = "6",
doi = "10.1121/1.4874358",
language = "English",
volume = "135",
pages = "EL284--EL290",
journal = "Journal of the Acoustical Society of America",
issn = "0001-4966",
publisher = "Acoustical Society of America",
number = "6",

}

Selection of spectral compressive operator for vector Taylor series-based model adaptation in noisy environments. / Baek, Soonho; Kang, Hong Goo.

In: Journal of the Acoustical Society of America, Vol. 135, No. 6, 06.2014, p. EL284-EL290.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Selection of spectral compressive operator for vector Taylor series-based model adaptation in noisy environments

AU - Baek, Soonho

AU - Kang, Hong Goo

PY - 2014/6

Y1 - 2014/6

N2 - This letter investigates the impact of spectral compression on the vector Taylor series-based model adaptation algorithm. Unlike mel-frequency cepstral coefficients obtained by the logarithmic compression, the fractional power compression is used for extracting features. Since the relationship between acoustic models for clean and noisy speech depends on nonlinearity of the spectrum, it is important to select an appropriate compressive operator in the model adaptation. In this letter, the dependency of spectral nonlinearity on the speech recognition system is analyzed in various noisy environments. Experimental results confirm that the replacement of the compressive operator improves the performance of the model adaptation.

AB - This letter investigates the impact of spectral compression on the vector Taylor series-based model adaptation algorithm. Unlike mel-frequency cepstral coefficients obtained by the logarithmic compression, the fractional power compression is used for extracting features. Since the relationship between acoustic models for clean and noisy speech depends on nonlinearity of the spectrum, it is important to select an appropriate compressive operator in the model adaptation. In this letter, the dependency of spectral nonlinearity on the speech recognition system is analyzed in various noisy environments. Experimental results confirm that the replacement of the compressive operator improves the performance of the model adaptation.

UR - http://www.scopus.com/inward/record.url?scp=84903138188&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84903138188&partnerID=8YFLogxK

U2 - 10.1121/1.4874358

DO - 10.1121/1.4874358

M3 - Article

AN - SCOPUS:84903138188

VL - 135

SP - EL284-EL290

JO - Journal of the Acoustical Society of America

JF - Journal of the Acoustical Society of America

SN - 0001-4966

IS - 6

ER -