Speaker adaptation algorithms based on piece-wise moving adaptive segment quantization method

Yoshinao Shiraki, Masaaki Honda

Research output: Chapter in Book/Report/Conference proceedingConference contribution

6 Citations (Scopus)

Abstract

A speaker adaptation method is presented based on a piece wise linear mapping of spectral code vector space into the spectral vector space of an unknown speaker. Adaptation is performed by modifying the code vectors to give a better fit to input the spectral vectors, while maintaining the local continuity of distribution of the original code vectors. Two adaptation algorithms are presented: a minimum-distortion method which provides a monotonic nonincreasing distortion for the training data and a simplified method. These algorithms are investigated in terms of distortion performance and adaptation speed. Experimental results show that the method provides much higher adaptation speed than the LBG (Linde-Buzo-Gray) algorithm and that the distortion performance approaches that of the LBG algorithm as the size of the training data increases.

Original languageEnglish
Title of host publicationICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
PublisherPubl by IEEE
Pages657-660
Number of pages4
Volume2
Publication statusPublished - 1990
Externally publishedYes
Event1990 International Conference on Acoustics, Speech, and Signal Processing: Speech Processing 2, VLSI, Audio and Electroacoustics Part 2 (of 5) - Albuquerque, New Mexico, USA
Duration: 1990 Apr 31990 Apr 6

Other

Other1990 International Conference on Acoustics, Speech, and Signal Processing: Speech Processing 2, VLSI, Audio and Electroacoustics Part 2 (of 5)
CityAlbuquerque, New Mexico, USA
Period90/4/390/4/6

Fingerprint

Vector spaces
vector spaces
education
continuity

ASJC Scopus subject areas

  • Signal Processing
  • Electrical and Electronic Engineering
  • Acoustics and Ultrasonics

Cite this

Shiraki, Y., & Honda, M. (1990). Speaker adaptation algorithms based on piece-wise moving adaptive segment quantization method. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings (Vol. 2, pp. 657-660). Publ by IEEE.

Speaker adaptation algorithms based on piece-wise moving adaptive segment quantization method. / Shiraki, Yoshinao; Honda, Masaaki.

ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Vol. 2 Publ by IEEE, 1990. p. 657-660.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Shiraki, Y & Honda, M 1990, Speaker adaptation algorithms based on piece-wise moving adaptive segment quantization method. in ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. vol. 2, Publ by IEEE, pp. 657-660, 1990 International Conference on Acoustics, Speech, and Signal Processing: Speech Processing 2, VLSI, Audio and Electroacoustics Part 2 (of 5), Albuquerque, New Mexico, USA, 90/4/3.
Shiraki Y, Honda M. Speaker adaptation algorithms based on piece-wise moving adaptive segment quantization method. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Vol. 2. Publ by IEEE. 1990. p. 657-660
Shiraki, Yoshinao ; Honda, Masaaki. / Speaker adaptation algorithms based on piece-wise moving adaptive segment quantization method. ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Vol. 2 Publ by IEEE, 1990. pp. 657-660
@inproceedings{ff7e3f8b1e16403b9dd84430c6c75c52,
title = "Speaker adaptation algorithms based on piece-wise moving adaptive segment quantization method",
abstract = "A speaker adaptation method is presented based on a piece wise linear mapping of spectral code vector space into the spectral vector space of an unknown speaker. Adaptation is performed by modifying the code vectors to give a better fit to input the spectral vectors, while maintaining the local continuity of distribution of the original code vectors. Two adaptation algorithms are presented: a minimum-distortion method which provides a monotonic nonincreasing distortion for the training data and a simplified method. These algorithms are investigated in terms of distortion performance and adaptation speed. Experimental results show that the method provides much higher adaptation speed than the LBG (Linde-Buzo-Gray) algorithm and that the distortion performance approaches that of the LBG algorithm as the size of the training data increases.",
author = "Yoshinao Shiraki and Masaaki Honda",
year = "1990",
language = "English",
volume = "2",
pages = "657--660",
booktitle = "ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",
publisher = "Publ by IEEE",

}

TY - GEN

T1 - Speaker adaptation algorithms based on piece-wise moving adaptive segment quantization method

AU - Shiraki, Yoshinao

AU - Honda, Masaaki

PY - 1990

Y1 - 1990

N2 - A speaker adaptation method is presented based on a piece wise linear mapping of spectral code vector space into the spectral vector space of an unknown speaker. Adaptation is performed by modifying the code vectors to give a better fit to input the spectral vectors, while maintaining the local continuity of distribution of the original code vectors. Two adaptation algorithms are presented: a minimum-distortion method which provides a monotonic nonincreasing distortion for the training data and a simplified method. These algorithms are investigated in terms of distortion performance and adaptation speed. Experimental results show that the method provides much higher adaptation speed than the LBG (Linde-Buzo-Gray) algorithm and that the distortion performance approaches that of the LBG algorithm as the size of the training data increases.

AB - A speaker adaptation method is presented based on a piece wise linear mapping of spectral code vector space into the spectral vector space of an unknown speaker. Adaptation is performed by modifying the code vectors to give a better fit to input the spectral vectors, while maintaining the local continuity of distribution of the original code vectors. Two adaptation algorithms are presented: a minimum-distortion method which provides a monotonic nonincreasing distortion for the training data and a simplified method. These algorithms are investigated in terms of distortion performance and adaptation speed. Experimental results show that the method provides much higher adaptation speed than the LBG (Linde-Buzo-Gray) algorithm and that the distortion performance approaches that of the LBG algorithm as the size of the training data increases.

UR - http://www.scopus.com/inward/record.url?scp=0025593309&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0025593309&partnerID=8YFLogxK

M3 - Conference contribution

VL - 2

SP - 657

EP - 660

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

PB - Publ by IEEE

ER -