Adaptation of Hidden Markov models using maximum model distance algorithm

Q. H. HE; S. KWONG; Q. Y. HONG

doi:10.1109/TSMCA.2003.818884

Adaptation of Hidden Markov models using maximum model distance algorithm

Q. H. HE^*, S. KWONG^*, Q. Y. HONG

^*Corresponding author for this work

Research output: Journal Publications › Comment / Debate › Research

5 Citations (Scopus)

Abstract

This paper presents a new approach that uses the maximum model distance (MMD) method for the adaptation of Hidden Markov models (HMMs). This method has the same framework as it is used for constructing speech recognizers with abundant data, and work effectively with any amount of adaptation data. All parameters of the HMMs with or without the adaptation data could be adapted. If the adaptation data is sufficient, then the adapted models will gradually become a speaker-dependent one. Both the dialect and the speaker adaptation experiments were conducted to investigate the effectiveness of the proposed algorithm. In the speaker adaptation experiments, up to 65.55% phoneme error reduction was achieved, and the MMD could reduce the phoneme error by 16.91% even only one adaptation utterance is available.

Original language	English
Pages (from-to)	270-276
Number of pages	7
Journal	IEEE Transactions on Systems, Man, and Cybernetics Part A:Systems and Humans.
Volume	34
Issue number	2
Early online date	26 Feb 2004
DOIs	https://doi.org/10.1109/TSMCA.2003.818884
Publication status	Published - Mar 2004
Externally published	Yes

Bibliographical note

Funding Information:
This work was supported in part by the City University of Hong Kong under Grant 7001488.

Keywords

Hidden Markov model
Maximum model distance
Speaker adaptation

Access to Document

10.1109/TSMCA.2003.818884

Cite this

@article{a5e39476dc2b49698a37d9aa2613f193,

title = "Adaptation of Hidden Markov models using maximum model distance algorithm",

abstract = "This paper presents a new approach that uses the maximum model distance (MMD) method for the adaptation of Hidden Markov models (HMMs). This method has the same framework as it is used for constructing speech recognizers with abundant data, and work effectively with any amount of adaptation data. All parameters of the HMMs with or without the adaptation data could be adapted. If the adaptation data is sufficient, then the adapted models will gradually become a speaker-dependent one. Both the dialect and the speaker adaptation experiments were conducted to investigate the effectiveness of the proposed algorithm. In the speaker adaptation experiments, up to 65.55% phoneme error reduction was achieved, and the MMD could reduce the phoneme error by 16.91% even only one adaptation utterance is available.",

keywords = "Hidden Markov model, Maximum model distance, Speaker adaptation",

author = "HE, {Q. H.} and S. KWONG and HONG, {Q. Y.}",

note = "Funding Information: This work was supported in part by the City University of Hong Kong under Grant 7001488. ",

year = "2004",

month = mar,

doi = "10.1109/TSMCA.2003.818884",

language = "English",

volume = "34",

pages = "270--276",

journal = "IEEE Transactions on Systems, Man, and Cybernetics Part A:Systems and Humans.",

issn = "1083-4427",

publisher = "IEEE Advancing Technology for Humanity",

number = "2",

}

TY - JOUR

T1 - Adaptation of Hidden Markov models using maximum model distance algorithm

AU - HE, Q. H.

AU - KWONG, S.

AU - HONG, Q. Y.

N1 - Funding Information: This work was supported in part by the City University of Hong Kong under Grant 7001488.

PY - 2004/3

Y1 - 2004/3

N2 - This paper presents a new approach that uses the maximum model distance (MMD) method for the adaptation of Hidden Markov models (HMMs). This method has the same framework as it is used for constructing speech recognizers with abundant data, and work effectively with any amount of adaptation data. All parameters of the HMMs with or without the adaptation data could be adapted. If the adaptation data is sufficient, then the adapted models will gradually become a speaker-dependent one. Both the dialect and the speaker adaptation experiments were conducted to investigate the effectiveness of the proposed algorithm. In the speaker adaptation experiments, up to 65.55% phoneme error reduction was achieved, and the MMD could reduce the phoneme error by 16.91% even only one adaptation utterance is available.

AB - This paper presents a new approach that uses the maximum model distance (MMD) method for the adaptation of Hidden Markov models (HMMs). This method has the same framework as it is used for constructing speech recognizers with abundant data, and work effectively with any amount of adaptation data. All parameters of the HMMs with or without the adaptation data could be adapted. If the adaptation data is sufficient, then the adapted models will gradually become a speaker-dependent one. Both the dialect and the speaker adaptation experiments were conducted to investigate the effectiveness of the proposed algorithm. In the speaker adaptation experiments, up to 65.55% phoneme error reduction was achieved, and the MMD could reduce the phoneme error by 16.91% even only one adaptation utterance is available.

KW - Hidden Markov model

KW - Maximum model distance

KW - Speaker adaptation

UR - http://www.scopus.com/inward/record.url?scp=1542500506&partnerID=8YFLogxK

U2 - 10.1109/TSMCA.2003.818884

DO - 10.1109/TSMCA.2003.818884

M3 - Comment / Debate

AN - SCOPUS:1542500506

SN - 1083-4427

VL - 34

SP - 270

EP - 276

JO - IEEE Transactions on Systems, Man, and Cybernetics Part A:Systems and Humans.

JF - IEEE Transactions on Systems, Man, and Cybernetics Part A:Systems and Humans.

IS - 2

ER -

Adaptation of Hidden Markov models using maximum model distance algorithm

Abstract

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this