FIELD: information technologies.
SUBSTANCE: prior to assessment of subband masking thresholds, the processed voice frame is identified for reference to one of the finished number of non-crossing classes M, then one of M units of multiband code books is selected, and then subband masking thresholds are assessed, at the same time number of subband code books and number of levels in these code books in M units of multiband code books differs depending in type of a processed voice frame and perceptual importance of each analysed subbands of a segment of a broadband voice signal. Having identified the value of perceptual entropy, a vector is formed, which contains number of the selected unit of multiband code books and depth of search in subband code books of this unit, and then, using the value of index of excitation vector a signal of excitation is selected from the selected unit of multiband code books, at the same time the output bit sequence of the coder is generated from the idenx of the excitation vector, quantised value of its amplification ratio, index of quantised vector, containing a number of the selected unit of multiband code books and depth of search in subband code books of this unit, and also a quantised vector of coefficients of linear forecasting.
EFFECT: increased quality of recovered voice messages at fixed speed of transmission.
3 dwg
