Continuous speech recognition using segmental unit input HMMs with a mixture of probability density functions and context dependency

It is well-known that HMMs only of the basic structure cannot capture the correlations among successive frames adequately. In our previous work, to solve this problem, segmental unit HMMs were introduced and their e ectiveness was shown. And the integration of cepstrum and cepstrum into the segmental unit HMMs was also found to improve the recognition performance in the work. In this paper, we investigated further re nements of the models by using a mixture of PDFs and/or context dependency, where, for a given syllable, only a preceding vowel was treated as the context information. Recognition experiments showed that the accuracy rate was improved by 23 %, which clearly indicates the e ectiveness of the renements examined in this paper. The proposed syllablebased HMM outperformed a triphone model.