5th International Conference on Spoken Language Processing

Sydney, Australia
November 30 - December 4, 1998

Maximum-Likelihood Updates Of HMM Duration Parameters For Discriminative Continuous Speech Recognition

Rathinavelu Chengalvarayan

Lucent Technologies, USA

Previous studies showed that a significantly enhanced recognition performance can be achieved by incorporating information about HMM duration along with the cepstral parameters. The reestimation formula for the duration parameters have been derived in the past using fixed segmentation during K-means training and the duration statistics are always fixed throughout the additional minimum string error (MSE) training process. In this study, we update the duration parameters along with other model parameters during discriminative training iterations. The convergence property of the training property based on the MSE approach is investigated, and experimental results on wireline connected digit recognition task demonstrated a 6% word error rate reduction by using the newly trained duration model parameters as compared to fixed duartion parameters during MSE training.

Full Paper

Bibliographic reference.  Chengalvarayan, Rathinavelu (1998): "Maximum-likelihood updates of HMM duration parameters for discriminative continuous speech recognition", In ICSLP-1998, paper 0021.