A NEW APPROACH TO GENERALIZED MIXTURE TYING FOR CONTINUOUS HMM-BASED SPEECH RECOGNITION

Daniel Willett; Gerhard Rigoll

A NEW APPROACH TO GENERALIZED MIXTURE TYING FOR CONTINUOUS HMM-BASED SPEECH RECOGNITION

Daniel Willett, Gerhard Rigoll

University of Duisburg-Essen

Research output: Contribution to conference › Paper › peer-review

6 Scopus citations

Abstract

In this paper we present a new approach for a generalized tying of mixture components for continuous mixture-density HMM-based speech recognition systems. With an iterative pruning and splitting procedure for the mixture components, this approach offers a very accurate and detailed representation of the acoustic space and at the same time keeps the number of parameters reasonably small in favor of a robust parameter estimation and a fast decoding. Contrary to other approaches, it does not require a strict clustering of the pdfs into subsets that share their mixture components, so that it is capable of providing more general and flexible types of mixture tying. We applied the new approach on a semi-continuous HMM (SCHMM)-system for the Resource Management task and improved its recognition performance by 12% and vastly accelerated the decoding because of a much faster likelihood computation.

Original language	English
Pages	1175-1178
Number of pages	4
State	Published - 1997
Externally published	Yes
Event	5th European Conference on Speech Communication and Technology, EUROSPEECH 1997 - Rhodes, Greece Duration: 22 Sep 1997 → 25 Sep 1997

Conference

Conference	5th European Conference on Speech Communication and Technology, EUROSPEECH 1997
Country/Territory	Greece
City	Rhodes
Period	22/09/97 → 25/09/97

Cite this

@conference{cca5d31818c54f439a83a3b0f3e19f61,

title = "A NEW APPROACH TO GENERALIZED MIXTURE TYING FOR CONTINUOUS HMM-BASED SPEECH RECOGNITION",

abstract = "In this paper we present a new approach for a generalized tying of mixture components for continuous mixture-density HMM-based speech recognition systems. With an iterative pruning and splitting procedure for the mixture components, this approach offers a very accurate and detailed representation of the acoustic space and at the same time keeps the number of parameters reasonably small in favor of a robust parameter estimation and a fast decoding. Contrary to other approaches, it does not require a strict clustering of the pdfs into subsets that share their mixture components, so that it is capable of providing more general and flexible types of mixture tying. We applied the new approach on a semi-continuous HMM (SCHMM)-system for the Resource Management task and improved its recognition performance by 12% and vastly accelerated the decoding because of a much faster likelihood computation.",

author = "Daniel Willett and Gerhard Rigoll",

note = "Publisher Copyright: {\textcopyright} 1997 5th European Conference on Speech Communication and Technology, EUROSPEECH 1997. All rights reserved.; 5th European Conference on Speech Communication and Technology, EUROSPEECH 1997 ; Conference date: 22-09-1997 Through 25-09-1997",

year = "1997",

language = "English",

pages = "1175--1178",

}

TY - CONF

T1 - A NEW APPROACH TO GENERALIZED MIXTURE TYING FOR CONTINUOUS HMM-BASED SPEECH RECOGNITION

AU - Willett, Daniel

AU - Rigoll, Gerhard

PY - 1997

Y1 - 1997

N2 - In this paper we present a new approach for a generalized tying of mixture components for continuous mixture-density HMM-based speech recognition systems. With an iterative pruning and splitting procedure for the mixture components, this approach offers a very accurate and detailed representation of the acoustic space and at the same time keeps the number of parameters reasonably small in favor of a robust parameter estimation and a fast decoding. Contrary to other approaches, it does not require a strict clustering of the pdfs into subsets that share their mixture components, so that it is capable of providing more general and flexible types of mixture tying. We applied the new approach on a semi-continuous HMM (SCHMM)-system for the Resource Management task and improved its recognition performance by 12% and vastly accelerated the decoding because of a much faster likelihood computation.

AB - In this paper we present a new approach for a generalized tying of mixture components for continuous mixture-density HMM-based speech recognition systems. With an iterative pruning and splitting procedure for the mixture components, this approach offers a very accurate and detailed representation of the acoustic space and at the same time keeps the number of parameters reasonably small in favor of a robust parameter estimation and a fast decoding. Contrary to other approaches, it does not require a strict clustering of the pdfs into subsets that share their mixture components, so that it is capable of providing more general and flexible types of mixture tying. We applied the new approach on a semi-continuous HMM (SCHMM)-system for the Resource Management task and improved its recognition performance by 12% and vastly accelerated the decoding because of a much faster likelihood computation.

UR - http://www.scopus.com/inward/record.url?scp=85017305906&partnerID=8YFLogxK

M3 - Paper

AN - SCOPUS:85017305906

SP - 1175

EP - 1178

T2 - 5th European Conference on Speech Communication and Technology, EUROSPEECH 1997

Y2 - 22 September 1997 through 25 September 1997

ER -

A NEW APPROACH TO GENERALIZED MIXTURE TYING FOR CONTINUOUS HMM-BASED SPEECH RECOGNITION

Abstract

Conference

Other files and links

Fingerprint

Cite this