The trended HMM with discriminative training for phonetic classification


In this paper, we extend the Maximum Likelihood (ML) training algorithm to the Minimum classificnrtion Emor (MCE) training algorithm for optimal estimation of the state-dependent polynomial coefficients in the trended HMM [2]. The problem of automatic speech recognition is viewed as a discriminative dynamic data-fitting problem, where nlotiue (not absolute) closeness in fitting an array of dynamic speech models to the unknown speech data sequence provides the recognition decision. In this view, the properties of the MCE formulation for training the trended HMM are analyzed by fitting raw s p d data using MCEtrained trended BMMS, contrasting the poor discriminative fitting using the MIrtrained models. Comparisons between the phonetic clasjsification as well as data-fitting results obtained with ML and -114th MCE training algorithms demonstrate the effectiveness of the discriminatively trained trended HMMs.

Extracted Key Phrases

3 Figures and Tables

Cite this paper

@inproceedings{Chengalvarayan1996TheTH, title={The trended HMM with discriminative training for phonetic classification}, author={Rathinavelu Chengalvarayan and Li Deng}, booktitle={ICSLP}, year={1996} }