A new class of fenonic Markov word models for large vocabulary continuous speech recognition

Published 1 January 1991

L.R. Bahl, J.R. Bellegarda, P.V. deSouza, P.S. Gopalakrishnan, D. Nahamoo, Michael Picheny

DOI10.1109/icassp.1991.150306

Citations10

TL;DR

Experimental results obtained on a 5000-word vocabulary continuous speech recognition task are presented to illustrate some of the benefits associated with the new models, which are more flexible than previously reported fenone-based word models and lead to an improved capability of modeling variations in pronunciation.

Abstract

A technique for constructing hidden Markov models for the acoustic representation of words is described. The models, built from combinations of acoustically based subword units called fenones, are derived automatically from one or more sample utterances of words. They are more flexible than previously reported fenone-based word models and lead to an improved capability of modeling variations in pronunciation. In addition, their construction is simplified, because it can be done using the well-known forward-backward algorithm for the parameter estimation of hidden Markov models. Experimental results obtained on a 5000-word vocabulary continuous speech recognition task are presented to illustrate some of the benefits associated with the new models. Multonic baseforms resulted in a reduction of 16% in the average error rate obtained for ten speakers.>

Keywords

Computer Science

IEEE Transactions on Pattern Analysis and Machine IntelligenceA Maximum Likelihood Approach to Continuous Speech Recognition

1,220 Citations1983L.R. Bahl, Frederick Jelinek +1 more

This paper describes a number of statistical models for use in speech recognition, with special attention to determining the parameters for such models from sparse data, and describes two decoding methods appropriate for constrained artificial languages and one appropriate for more realistic decoding tasks.

Context-dependent modeling for acoustic-phonetic recognition of continuous speech

218 Citations2005Richard Schwartz, Y.-L. Chow +4 more

The combination of general spectral information and specific acoustic-phonetic features is shown to result in more accurate phonetic recognition than either representation by itself.

Context-Dependent Phonetic Hidden Markov Models for Speaker-Independent Continuous Speech Recognition

213 Citations1992K.-F. Lee

Two new context-dependent phonetic units are introduced: function-word-dependent phone models, which focus on the most difficult subvocabulary; and generalized triphones, which combine similar triphones on the basis of an information-theoretic measure.

IEEE Transactions on Acoustics Speech and Signal ProcessingTied mixture continuous parameter modeling for speech recognition

157 Citations1990J.R. Bellegarda, D. Nahamoo

A class of very general hidden Markov models which can accommodate feature vector sequences lying either in a discrete or in a continuous space is considered; the new class allows one to represent the prototypes in an assumption-limited, yet convenient way, as tied mixtures of simple multivariate densities.

Decision trees for phonological rules in continuous speech

109 Citations1991L.R. Bahl, P.V. deSouza +3 more

An automatic method for modeling phonological variation using decision trees that specifies the acoustic realization of the phone as a function of the context in which it appears is presented.

IEEE Transactions on Acoustics Speech and Signal ProcessingThe role of phonological rules in speech understanding research

98 Citations1975Beatrice T. Oshika, Victor W. Zue +3 more

Acoustic Markov models used in the Tangora speech recognition system

79 Citations2003L.R. Bahl, Peter F. Brown +2 more

An automatic technique for constructing Markov word models is described and results are included of experiments with speaker-dependent and speaker-independent models on several isolated-word recognition tasks.

International Conference on Acoustics, Speech, and Signal ProcessingLarge vocabulary natural language continuous speech recognition

65 Citations2003L.R. Bahl, R. Bakis +12 more

A description is presented of the authors' current research on automatic speech recognition of continuously read sentences from a naturally-occurring corpus: office correspondence, which combines features from their current isolated-word recognition system and from their previously developed continuous-speech recognition system.

1st IASTED International Symposium on Signal Processing and its ApplicationsAutomatic construction of acoustic markov models for words

7 Citations1987LR Bahl, P. F. Brown +3 more

An automatic technique for constructing Markov word models from sub-word units called fenones, which are represented by miniature Markov models, are very short speech events and may be obtained automatically via vector quantization.