PREDICTION BASED ON MODEL IN FILTER SET WITH CRITICAL DISCRETEIZATION Russian patent published in 2017 - IPC G10L19/26 G10L19/24 G10L19/93 G10L19/02 

FIELD: physics.

SUBSTANCE: signal of the first audio subband is determined using a set of analyzing filters comprising a series of analyzing filters that generate a series of subband signals from the audio signal, respectively, in a row of subbands. A method for estimating the first discrete signal value of the first subband signal in the first subband of an audio signal includes determining a model parameter for a signal model; determining a prediction coefficient to be applied to the previous discrete value of the decoded signals of the first subband obtained from the first subband signal based on the signal model based on the model parameter and based on the set of analyzing filters. The time interval of the previous discrete value precedes the time interval of the first discrete value; and determining the estimation of the first discrete value by applying the prediction coefficient to the previous discrete value.

EFFECT: providing a low data rate with a low level of spurious frequencies.

33 cl, 9 dwg