MARKOV MODELS IN THE PROBLEM OF THE SPEECH PARAMETRIZATION

 

D. E. Prozorov, K. V. Pletnev

Federal Government-financed Educational Institution of Higher Professional Education

«Vyatka State University» (FGEI HPE “VyatSU”)

Federal Government-financed Educational Institution of Higher Professional Education

“Vyatka State University of Humanities”

 

The paper is received on June 2, 2014

 

Abstract: The article investigates the possibility of using of the speech parameterization methods by simple and complex (multivariable) Markov chains in systems of automatic recognition of commands (ARC). It is shown that the speech parametrization by multivariable Markov chains allows to reach a compromise between the efficiency of speech recognition commands and computational complexity of the algorithm parameterization. Low computational complexity Markov parameterization indicate the possibility of a significant reduction in requirements for computational complexity of ARC.

Key words: speech recognition, parameterization, multivariate Markov chain.