Discriminative training of GMM-HMM acoustic model by RPCL learning

Abstract

This paper presents a new discriminative approach for training Gaussian mixture models (GMMs) of hidden Markov models (HMMs) based acoustic model in a large vocabulary continuous speech recognition (LVCSR) system. This approach is featured by embedding a rival penalized competitive learning (RPCL) mechanism on the level of hidden Markov states. For every input, the correct identity state, called winner and obtained by the Viterbi force alignment, is enhanced to describe this input while its most competitive rival is penalized by de-learning, which makes GMMs-based states become more discriminative.Without the extensive computing burden required by typical discriminative learning methods for one-pass recognition of the training set, the new approach saves computing costs considerably. Experiments show that the proposed method has a good convergence with better performances than the classical maximum likelihood estimation (MLE) based method. Comparing with two conventional discriminative methods, the proposed method demonstrates improved generalization ability, especially when the test set is not well matched with the training set.

Keywords

discriminative training / hidden Markov model / rival penalized competitive learning / Bayesian Ying-Yang harmony learning / large vocabulary continuous speech recognition

References

[1] Brown P. The acoustic-modeling problem in automatic speech recognition. Dissertation for the Doctoral Degree. Pittsburgh: Carnegie Mellon University, 1987

[2] Gales M, Young S. The application of hidden Markov models in speech recognition. Foundations and Trends in Signal Processing, 2008, 1(3): 195-304

[3] Bahl L, Brown P, De Souza P, Mercer R. Maximum mutual information estimation of hidden Markov model parameters for speech recognition. In: Proceedings of 1986 IEEE International Conference on Acoustics, Speech, and Signal Processing. 1986, 11: 49-52

[4] Juang B H, Chou W, Lee C H. Minimum classification error rate methods for speech recognition. IEEE Transactions on Speech and Audio Processing, 1997, 5(3): 257-265

[5] Povey D, Woodland P C. Minimum phone error and Ismothing for improved discriminative training. In: Proceedings of 2002 IEEE International Conference on Acoustics, Speech, and Signal Processing. 2002, 1: 105-108

[6] Su D, Wu X H, Xu L. GMM-HMM acoustic model training by a two level procedure with Gaussian components determined by automatic model selection. In: Proceedings of 2010 IEEE International Conference on Acoustics, Speech, and Signal Processing. 2010, 4890-4893

[7] Xu L. Bayesian Ying-Yang system, best harmony learning, and five action circling. Frontiers of Electrical and Electronic Engineering in China, 2010, 5(3): 281-328

[8] Xu L, Krzyzak A, Oja E. Unsupervised and supervised classifications by rival penalized competitive learning. In: Proceedings of the 11th International Conference on Pattern Recognition. 1992, II: 672-675

[9] Xu L. Rival penalized competitive learning. Scholarpedia, 2007, 2(8): 1810

[10] Xu L. A unified perspective and new results on RHT computing, mixture based learning, and multi-learner based problem solving. Pattern Recognition, 2007, 40(8): 2129-2153

[11] Kuhn H W. The hungarian method for the assignment problem. Naval Research Logistics Quarterly, 1955, 2(1-2): 83-97

[12] Young S, Kershaw D, Odell J, Ollason D, Valtchev V, Woodland P. The HTK Book Version 3.4. Cambridge: Cambridge University Press, 2006

[13] Povey D, Kingsbury B. Evaluation of proposed modifications to MPE for large scale discriminative training. In: Proceedings of 2007 IEEE International Conference on Acoustics, Speech, and Signal Processing. 2007, 4: IV-321-IV-324

[14] Cheng Y J, Lin C K, Lee L S. Evaluation and analysis of minimum phone error training and its modified versions for large vocabulary Mandarin speech recognition. In: Proceedings of 2008 IEEE International Symposium on Chinese Spoken Language Processing. 2008, 1: 157-160

[15] Valtchev V, Odell J J, Woodland P C, Young S J. MMIE training of large vocabulary recognition systems. Speech Communication, 1997, 22(4): 303-314

[16] McDermott E, Katagiri S. String-level MCE for continuous phoneme recognition. In: Proceedings of EuroSpeech 1997. 1997, 123-126

[17] Macherey W, Haferkamp L, Schluter R, Ney H. Investigations on error minimizing training criteria for discriminative training in acoustic speech recognition. In: Proceedings of EuroSpeech 2005. 2005, 2133-2136

[18] Schlter R, Macherey W, Mller B, Ney H. Comparison of discriminative training criteria and optimization methods for speech recognition. Speech Communication, 2001, 34(3): 287-310

[19] Fu Q, He X, Deng L. Phone-discriminating minimum classification error (P-MCE) training criteria for phonetic recognition. In: Proceedings of InterSpeech 2007. 2007, 2073-2076

RIGHTS & PERMISSIONS

2014 Higher Education Press and Springer-Verlag Berlin Heidelberg