Comparison of parameters for the selection and classification of voiced phonemes segments

Abstract: The applicability of the parameters for the speech signal analysis to extract voiced segments, and to distinguish phonemes was investigated. The best option in case of application as a crucial block of a neural network – a two-layer perceptron – feeding the input channel number with the maximum intensity and the correlation between samples within the analysis window was compared. More informative parameter of the correlation between samples in comparison with the number of channels with a maximum intensity for the solution of the problem of finding sites and voiced the inapplicability of both parameters for the classification of phonemes was shown.

Keywords: speech analysis, neural network, the selection of segments

For citation:
Elistratov S. A., Kosenko M. A., Kostyuchenko E. Yu., Chicherin A. A. Comparison of parameters for the selection and classification of voiced phonemes segments. Doklady Tomskogo gosudarstvennogo universiteta sistem upravleniya i radioelektroniki, 2012, no. 1(25), – p. 2. pp. 171–174.