Explicit Modelling of State Duration Correlations in Hidden Markov Models

Abstract

In recent years considerable effort has been directed towards improving the treatment of durational structure in hidden Markov model (HMM) based approaches to speech pattern modelling. In general these studies have been concerned with more accurate modelling of the variations in segment duration which occur when words are spoken at a nominally constant speaking rate. However, recent work has shown that some of the performance gains which can be achieved by improved duration modelling are lost when the words in the test set are spoken at a different rate from those in the training set. This memorandum presents an approach to solving this problem based on the capture and use of information about state duration correlations. A method for measuring correlations between the durations of adjacent states in a HMM is described. The method involves expanding a standard HMM into a special type of hidden semi-Markov model (HSMM), called a Correlated Duration HMM (CDHMM), in which each state of the original HMM is expanded into a set of fixed-duration HSMM states. The probabilities associated with transitions between these states are measures of state duration correlation. Experiments are described in which the CDHMM method is applied to a set of sentences spoken at four different speaking rates. Great Britain.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Sep 01, 1988
Accession Number
ADA203174

Entities

People

  • L. Sime
  • M. J. Russell

Organizations

  • Royal Signals and Radar Establishment

Tags

Communities of Interest

  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Algorithms
  • Automated Speech Recognition
  • Data Science
  • Hidden Markov Models
  • Information Science
  • Markov Chains
  • Markov Models
  • Markov Processes
  • Models
  • Probabilistic Models
  • Probability
  • Recognition
  • Standards
  • Stationary Processes
  • Stochastic Processes
  • Training
  • Transitions

Readers

  • Computational Modeling and Simulation
  • Speech Processing/Speech Recognition.