Using frame correlation algorithm in a duration distribution based hidden Markov Model

摘要

The assumption of frme independence is a widely known weakness oftraditional hidden Markov model (HMM). In this paper, a framecorrelation algorithm based on the duration distribution Based hiddenMarkov model (DDBHMM) is proposed. In the algorithm, an AR model isused to depict The low pass effect of vocal tract from which stemsthe inertia leading to frame correlation. In the pre- Liminaryexperiment of middle vocabulary speaker dependent isolated wordrecognition, our frame cor- Relation algorithm outperforms the frameindependent one. The average error reduction is about 20/100.