FIELD: physics.
SUBSTANCE: result is achieved by introducing linear prediction parametres of a predictor into a differential vector quantiser with adaptation to the voice pitch: when classifying the frame of a speech signal as stationary vocalised, the prediction scheme with adaptation to the voice pitch is selected. Calculation of the prediction vector involves calculation of the prediction error vector through prediction with adaptation to the voice pitch. Selection of one of multiple scaling schemes involves selection of scaling coefficient equal to one.
EFFECT: reduced number of bits allocated for encoding linear prediction parametres for stationary vocalised frames of a speech signal with minimal quantisation error.
4 dwg
