FIELD: radio engineering.
SUBSTANCE: device comprises source (S) of input sound signal, memory device (MD), analyzing device (AD), producing device (PD) and synthesising device (SD). PD is arranged on the basis of characteristics analyser (CA) and correcting processor (CP). Switch (S) of training/operation mode is introduced, as well as input signal analyser (ISA). S is connected to input of S. MD is equipped with unit of audio records (U). Input/output of S is connected to input/output of ISA, and its output - to input of U. The first output of U data is connected to input of ISA, and the second output of U data - to AD input. ISA is arranged as providing for decay of input voice signal into sinusoid components of signal (Sig), noise components of signal (N) and residual components of signal (R) and is arranged with the possibility to generate sets of characteristic vectors and functions of conversion for each specified components and their transmission into MD. AD is arranged as providing for decay of input voice signal from U into Sig, N and R. CA and CP are arranged with the possibility to process specified components.
EFFECT: invention makes it possible to perform a song by user's voice, but in manner and with quality level of professional singer performance with minimisation of performance errors and improvement of its quality.
6 cl, 5 dwg
