MARKOV MODELS OF THE QUANTIZED SPEECH SIGNALS

UDC 004.934.2

K.P. Pletnyov, D.E. Prozorov

The article considers Markov models of the quantized speech signals (QSS). It is shown that correlation properties of the bit sequences obtained by the quantization of speech signals, allow it to apply the mathematical apparatus of simple or complex (multiply connected) finite Markov chains for the statistical description of the QSS. The method of parameterization based on Markov models of QSS reduces the performance requirements of systems of automatic speech recognition.

Keywords: Markov models of speech signals, Markov chains, parametrization.

Full text:
PletnevProzorov_3_16_1.pdf