Large vocabulary word recognition using context-dependent allophonic hidden Markov models☆

Abstract In this paper, we report our development of context-dependent allophonic hidden Markov models (HMMs) implemented in a 75 000-word speaker-dependent Gaussian-HMM recognizer. The context explored is the immediate left and/or right adjacent phoneme. To achieve reliable estimation of the model parameters, phonemes are grouped into classes based on their expected co-articulatory effects on neighboring phonemes. Only five separate preceding and following contexts are identified explicitly for each phoneme. By grouping the contexts we ensure that they occur frequently enough in the training data to allow reliable estimation of the parameters of the HMM representing the context-dependent units. Further improvement in the estimation reliability is obtained by tying the covariance matrices in the HMM output distributions across all contexts. Speech recognition experiments show that when a large amount of data (e.g. over 2500 words) is used to train context-dependent HMMs, the word recognition error rate is reduced by 33%, compared with the context-independent HMMs. For smaller amounts of training data the error reduction becomes less significant.

[1]  P. Mermelstein,et al.  Fast search strategy in a large vocabulary word recognizer , 1988 .

[2]  Lalit R. Bahl,et al.  Further results on the recognition of a continuously read natural corpus , 1980, ICASSP.

[3]  Raj Reddy,et al.  Large-vocabulary speaker-independent continuous speech recognition: the sphinx system , 1988 .

[4]  Lalit R. Bahl,et al.  A Maximum Likelihood Approach to Continuous Speech Recognition , 1983, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[5]  Anne-Marie Derouault,et al.  Context-dependent phonetic Markov models for large vocabulary speech recognition , 1987, ICASSP '87. IEEE International Conference on Acoustics, Speech, and Signal Processing.

[6]  B. Merialdo.  Speech recognition with very large size dictionary , 1987, ICASSP '87. IEEE International Conference on Acoustics, Speech, and Signal Processing.

[7]  D. B. Paul,et al.  Speaker stress-resistant continuous speech recognition , 1988, ICASSP-88., International Conference on Acoustics, Speech, and Signal Processing.

[8]  F. Jelinek,et al.  Continuous speech recognition by statistical methods , 1976, Proceedings of the IEEE.

[9]  Michael Picheny,et al.  Acoustic Markov models used in the Tangora speech recognition system , 1988, ICASSP-88., International Conference on Acoustics, Speech, and Signal Processing.

[10]  Richard M. Schwartz,et al.  Improved hidden Markov modeling of phonemes for continuous speech recognition , 1984, ICASSP.

[11]  L. R. Rabiner,et al.  Some properties of continuous hidden Markov model representations , 1985, AT&T Technical Journal.

[12]  Li Deng,et al.  A locus model of coarticulation in an HMM speech recognizer , 1989, International Conference on Acoustics, Speech, and Signal Processing,.

[13]  Louis A. Liporace,et al.  Maximum likelihood estimation for multivariate observations of Markov sources , 1982, IEEE Trans. Inf. Theory.

[14]  L. Baum,et al.  An inequality and associated maximization technique in statistical estimation of probabilistic functions of a Markov process , 1972 .

[15]  Li Deng,et al.  Use of vowel duration information in a large vocabulary word recognizer , 1989 .

[16]  E. A. Martin,et al.  Multi-style training for robust isolated-word speech recognition , 1987, ICASSP '87. IEEE International Conference on Acoustics, Speech, and Signal Processing.

[17]  Stan Davis,et al.  Comparison of Parametric Representations for Monosyllabic Word Recognition in Continuously Spoken Se , 1980 .

[18]  John Makhoul,et al.  BYBLOS: The BBN continuous speech recognition system , 1987, ICASSP '87. IEEE International Conference on Acoustics, Speech, and Signal Processing.

[19]  L. Baum,et al.  Growth transformations for functions on manifolds. , 1968 .

[20]  L. Deng,et al.  Modeling microsegments of stop consonants in a hidden Markov model based word recognizer , 1990 .

[21]  Frank K. Soong,et al.  High performance connected digit recognition, using hidden Markov models , 1988, ICASSP-88., International Conference on Acoustics, Speech, and Signal Processing.

[22]  L. R. Rabiner,et al.  An introduction to the application of the theory of probabilistic functions of a Markov process to automatic speech recognition , 1983, The Bell System Technical Journal.