An algorithm for the dynamic inference of hidden Markov models (DIHMM)

The DIHMM algorithm performs a robust estimation of the HMM topology and parameters. It allows a better control of the speech variability within each state of the HMM, yielding enhanced estimates. The DIHMM parameters (number of states, structure of the Gaussian mixture density functions, transition matrix) are obtained from the training data via probabilistic grammatical inference techniques welded in a Viterbi-like training framework. Experimental results on various databases indicate a global improvement of the recognition rates in adverse environments; the results averaged on three databases show an increase of 12.8% on raw data and 2.4% when using NSS (nonlinear spectral subtraction).<<ETX>>

[1]  Joseph Picone.  On modeling duration in context in speech recognition , 1989, International Conference on Acoustics, Speech, and Signal Processing,.

[2]  Renato De Mori,et al.  High performance connected digit recognition using codebook exponents , 1992, [Proceedings] ICASSP-92: 1992 IEEE International Conference on Acoustics, Speech, and Signal Processing.

[3]  Francisco Casacuberta,et al.  Learning the structure of HMM's through grammatical inference techniques , 1990, International Conference on Acoustics, Speech, and Signal Processing.

[4]  Michael Picheny,et al.  A new class of fenonic Markov word models for large vocabulary continuous speech recognition , 1991, [Proceedings] ICASSP 91: 1991 International Conference on Acoustics, Speech, and Signal Processing.

[5]  Jérôme Boudy,et al.  Experiments with a nonlinear spectral subtractor (NSS), Hidden Markov models and the projection, for robust speech recognition in cars , 1991, Speech Commun..

[6]  J. Boudy,et al.  Non-linear spectral subtraction (NSS) and hidden Markov models for robust speech recognition in car noise environments , 1992, [Proceedings] ICASSP-92: 1992 IEEE International Conference on Acoustics, Speech, and Signal Processing.

[7]  Mark A. Clements,et al.  On the phonetic structure of a large hidden Markov model , 1991, [Proceedings] ICASSP 91: 1991 International Conference on Acoustics, Speech, and Signal Processing.

[8]  Jérôme Boudy,et al.  Root homomorphic deconvolution schemes for speech processing in car noise environments , 1993, 1993 IEEE International Conference on Acoustics, Speech, and Signal Processing.

[9]  Biing-Hwang Juang,et al.  The segmental K-means algorithm for estimating parameters of hidden Markov models , 1990, IEEE Trans. Acoust. Speech Signal Process..

[10]  Michael G. Thomason,et al.  Dynamic Programming Inference of Markov Networks from Finite Sets of Sample Strings , 1986, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[11]  Xuedong Huang,et al.  Semi-continuous hidden Markov models for speech signals , 1990 .

[12]  King-Sun Fu,et al.  A Step Towards Unification of Syntactic and Statistical Pattern Recognition , 1986, IEEE Transactions on Pattern Analysis and Machine Intelligence.