Maximum mutual information training of a neural predictive-based HMM speech recognition system

A corrective training scheme based on the maximum mutual information (MMI) criterion is developed for training a neural predictive-based HMM (hidden Markov model) speech recognition system. The performance of the system on speech recognition tasks when trained with this technique is compared to its performance when trained using the maximum likelihood (ML) criterion. Preliminary results obtained indicate the superiority of ML training over MMI training for predictive-based models. This result is in agreement with earlier findings in the literature regarding direct classification models.<<ETX>>

[1]  Harvey F. Silverman,et al.  Neural networks, maximum mutual information training, and maximum likelihood training (speech recognition) , 1990, International Conference on Acoustics, Speech, and Signal Processing.

[2]  Lalit R. Bahl,et al.  Maximum mutual information estimation of hidden Markov model parameters for speech recognition , 1986, ICASSP '86. IEEE International Conference on Acoustics, Speech, and Signal Processing.

[3]  Li Deng,et al.  Neural-network architecture for linear and nonlinear predictive hidden Markov models: application to speech recognition , 1991, Neural Networks for Signal Processing Proceedings of the 1991 IEEE Workshop.

[4]  Ken-ichi Iso,et al.  Speaker-independent word recognition using a neural prediction model , 1990, International Conference on Acoustics, Speech, and Signal Processing.

[5]  Alex Waibel,et al.  Continuous speech recognition using linked predictive neural networks , 1991, [Proceedings] ICASSP 91: 1991 International Conference on Acoustics, Speech, and Signal Processing.

[6]  Esther Levin,et al.  Word recognition using hidden control neural architecture , 1990, International Conference on Acoustics, Speech, and Signal Processing.