CONTINUOUSLY VARIABLE DURATION HIDDEN MARKOV MODELS FOR SPEECH ANALYSIS.

Research output: Contribution to journalConference articlepeer-review

Abstract

A hidden Markov model is described for which the probability of occupancy of a state over the duration tau is governed by the gamma density, d( tau ). It is shown how to estimate the parameters of this model, and a proof of the correctness of the method is sketched. The performance of the algorithm is demonstrated in a test on simulated data. For the five-state, three-dimensisonal case tested, convergence to correct values from random initial values is obtained in 7-10 iterations.

Original languageEnglish (US)
Pages (from-to)1241-1244
Number of pages4
JournalICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
StatePublished - Dec 1 1986
Externally publishedYes

ASJC Scopus subject areas

  • Software
  • Signal Processing
  • Electrical and Electronic Engineering

Fingerprint Dive into the research topics of 'CONTINUOUSLY VARIABLE DURATION HIDDEN MARKOV MODELS FOR SPEECH ANALYSIS.'. Together they form a unique fingerprint.

Cite this