A training method for hidden Markov model with maximum model distance and genetic algorithm

Q. Y. HONG; S. KWONG

doi:10.1109/ICNNSP.2003.1279309

A training method for hidden Markov model with maximum model distance and genetic algorithm

Q. Y. HONG, S. KWONG

Research output: Book Chapters | Papers in Conference Proceedings › Conference paper (refereed) › Research › peer-review

5 Citations (Scopus)

Abstract

Maximum model distance (MMD) is a discriminative algorithm developed for training the whole HMM models. It differs from the traditional maximum-likelihood (ML) approach through comparing the likelihood against those similar utterances and maximizes their likelihood differences. Combined with MMD, this paper proposes a hybrid training method based on the genetic algorithm (GA). Experimental results from the TI46-Word alphabet database show that this algorithm has better performance than MMD. The reason is that the MMD algorithm is exploring only one local maximum in practice while the GA operations in the hybrid algorithm provide the ability to explore several local maximums and hopefully the global maximum. © 2003 IEEE.

Original language	English
Title of host publication	Proceedings of 2003 International Conference on Neural Networks and Signal Processing
Publisher	IEEE
Pages	465-468
Number of pages	4
ISBN (Print)	0780377028
DOIs	https://doi.org/10.1109/ICNNSP.2003.1279309
Publication status	Published - 2003
Externally published	Yes
Event	International Conference on Neural Networks and Signal Processing, 2003 - Liuyuan Hotel, Nanjing, China Duration: 14 Dec 2003 → 17 Dec 2003

Conference

Conference	International Conference on Neural Networks and Signal Processing, 2003
Country/Territory	China
City	Nanjing
Period	14/12/03 → 17/12/03

Access to Document

10.1109/ICNNSP.2003.1279309

Cite this

@inproceedings{8ba208aff1f64699aa3563fe355ab277,

title = "A training method for hidden Markov model with maximum model distance and genetic algorithm",

abstract = "Maximum model distance (MMD) is a discriminative algorithm developed for training the whole HMM models. It differs from the traditional maximum-likelihood (ML) approach through comparing the likelihood against those similar utterances and maximizes their likelihood differences. Combined with MMD, this paper proposes a hybrid training method based on the genetic algorithm (GA). Experimental results from the TI46-Word alphabet database show that this algorithm has better performance than MMD. The reason is that the MMD algorithm is exploring only one local maximum in practice while the GA operations in the hybrid algorithm provide the ability to explore several local maximums and hopefully the global maximum. {\textcopyright} 2003 IEEE.",

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

year = "2003",

doi = "10.1109/ICNNSP.2003.1279309",

language = "English",

isbn = "0780377028",

pages = "465--468",

booktitle = "Proceedings of 2003 International Conference on Neural Networks and Signal Processing",

publisher = "IEEE",

note = "International Conference on Neural Networks and Signal Processing, 2003 ; Conference date: 14-12-2003 Through 17-12-2003",

}

HONG, QY & KWONG, S 2003, A training method for hidden Markov model with maximum model distance and genetic algorithm. in Proceedings of 2003 International Conference on Neural Networks and Signal Processing. IEEE, pp. 465-468, International Conference on Neural Networks and Signal Processing, 2003, Nanjing, China, 14/12/03. https://doi.org/10.1109/ICNNSP.2003.1279309

TY - GEN

T1 - A training method for hidden Markov model with maximum model distance and genetic algorithm

AU - HONG, Q. Y.

AU - KWONG, S.

PY - 2003

Y1 - 2003

N2 - Maximum model distance (MMD) is a discriminative algorithm developed for training the whole HMM models. It differs from the traditional maximum-likelihood (ML) approach through comparing the likelihood against those similar utterances and maximizes their likelihood differences. Combined with MMD, this paper proposes a hybrid training method based on the genetic algorithm (GA). Experimental results from the TI46-Word alphabet database show that this algorithm has better performance than MMD. The reason is that the MMD algorithm is exploring only one local maximum in practice while the GA operations in the hybrid algorithm provide the ability to explore several local maximums and hopefully the global maximum. © 2003 IEEE.

AB - Maximum model distance (MMD) is a discriminative algorithm developed for training the whole HMM models. It differs from the traditional maximum-likelihood (ML) approach through comparing the likelihood against those similar utterances and maximizes their likelihood differences. Combined with MMD, this paper proposes a hybrid training method based on the genetic algorithm (GA). Experimental results from the TI46-Word alphabet database show that this algorithm has better performance than MMD. The reason is that the MMD algorithm is exploring only one local maximum in practice while the GA operations in the hybrid algorithm provide the ability to explore several local maximums and hopefully the global maximum. © 2003 IEEE.

UR - http://www.scopus.com/inward/record.url?scp=71649100322&partnerID=8YFLogxK

U2 - 10.1109/ICNNSP.2003.1279309

DO - 10.1109/ICNNSP.2003.1279309

M3 - Conference paper (refereed)

SN - 0780377028

SP - 465

EP - 468

BT - Proceedings of 2003 International Conference on Neural Networks and Signal Processing

PB - IEEE

T2 - International Conference on Neural Networks and Signal Processing, 2003

Y2 - 14 December 2003 through 17 December 2003

ER -

A training method for hidden Markov model with maximum model distance and genetic algorithm

Abstract

Conference

Access to Document

Other files and links

Fingerprint

Cite this