Online adaptation of HMMs to real-life conditions: A unified framework

IEEE Transactions on Speech and Audio Processing

Volumn 9, Issue 4, 2001, Pages 342-357

  Mokbel, Chafic ^a,b,c  

^a IEEE   (Lebanon)

^b ORANGE LABS   (France)

^c UNIVERSITY OF BALAMAND   (Lebanon)

Author keywords

Adaptive training; Blind online adaptation; Estimate maximize (EM); Hidden Markov models; Maximum a posteriori; Maximum likelihood; Recursive EM

Indexed keywords

GAUSSIAN DISTRIBUTIONS; HIDDEN MARKOV PROCESSES (HMM);

ADAPTIVE ALGORITHMS; MARKOV PROCESSES; MAXIMUM LIKELIHOOD ESTIMATION; RECURSIVE FUNCTIONS; STATISTICAL METHODS;

SPEECH RECOGNITION;

EID: 0035340712     PISSN: 10636676     EISSN: None     Source Type: Journal    
DOI: 10.1109/89.917680     Document Type: Article

References (17)

1. Automatic word recognition in cars
2. A study on speaker adaptation of the parameters of continuous density hidden Markov models
3. Maximum a posteriori estimation for multivariate Gaussian mixture observations of Markov chains
4. Bayesian adaptation of speech recognizers to field speech data
5. Speaker adaptation of continuous density HMM's using multivariate linear regression
6. Adapting PSN recognition models to the GSM environment by using spectral transformation
7. On feature and model compensation approach to robust speech recognition
8. Toward improving ASR robustness for PSN and GSM telephone applications
9. MUSE: Multi-path stochastic equalization: A theoretical framework to combine equalization and stochastic modeling
10. Bayesian adaptive learning of the parameters of hidden Markov model for speech recognition
11. Maximum likelihood from incomplete data via the EM algorithm
12. Speaker adaptation using constrained estimation of Gaussian mixtures
13. Speaker adaptation with autonomous model complexity control by MDL principle
14. Telephone speech recognition based on Bayesian adaptation of hidden Markov models
15. On the modelization of allophones in a HMM based speech recognition system
16. Frame-synchronous adaptation of cepstrum by linear regression