FIELD: numeric coding of voice signals for their complex representation for transmission and storage. SUBSTANCE: proposed method based on vector quantization and de-quantization includes additional control by means of sound/noise signal used as control parameter T/III, T/III*, respectively. Speech converting device and alternatives of practical implementation of units incorporated in vector quantizer with finite number of states and vector de-quantizer with finite number of states provide for additional control of sound/noise signal in vector quantization and de-quantization procedure; provision is made for dividing set of reference code vectors into two subsets corresponding to vocalized and non-vocalized fragments of voice signal; this procedure raises quality of synthesized voice signal and provides for more precise description of voice signal transition sequence. EFFECT: improved intelligibility and reproducibility of synthesized voice signal. 5 cl, 10 dwg
