Markov Model of Quantized Speech Signal

Currently, automatic speech recognition systems are widely used. When developing such systems, significant attention is paid to choice of parameterization method of speech signals. The paper presents options for representing fragments of speech signals by simple and high-order Markov chains. Special cases of the implementation of these models are also presented. The results show that when using the methods of Markov parameterization in problems of automatic recognition of speech commands, it is advisable to use models of high order Markov chains. The considered models of speech signals can significantly reduce the requirements on the computing resources of automatic speech recognition systems.