FIELD: information technology.
SUBSTANCE: objective set in a linear prediction vocoder is achieved by avoiding transmission of drive signal information over a communication channel. The drive signal is identified immediately at reception from data on parameters of the synthesising model using a neural network. Information on coefficients of the forming model, gain factor, parameters which characterise the encoded speech signal which are calculated on each quasi-stationary analysis segment of the speech signal are transmitted over the communication channel.
EFFECT: low transmission rate of a low bit rate linear prediction vocoder with efficient encoding of speech signals, while maintaining quality factors of the synthesised signal.
3 dwg
