FIELD: speech recognition. SUBSTANCE: speech coding for adapting noise-like signals involves production of plurality of parameters from source voice signal wherefrom approximation of the latter can be recovered; to this end additional signal designed for presenting source voice signal is generated in response to this source voice signal; first difference is determined between signal waveform associated with source voice signal and signal waveform associated with additional signal; second difference is determined between energy parameter obtained from source voice signal and respective energy parameter associated with additional signal; then first and second differences are used to evaluate at least one of parameters that can be used to recover approximation of source voice signal. This process is implemented by speech coding device and communication system transceiver incorporating such coding device. EFFECT: improved coding of noise-like signals at reduced data transfer speeds without producing unwanted sounds during multimode coding. 27 cl, 7 dwg
