FIELD: coding/decoding of stereo audio signal. SUBSTANCE: method for coding audio signals into multilayer data stream that has main layer and at least two expansion layers involves processing of input audio signals and their quantizing for each predetermined coding range; then quantized data corresponding to main layer out of quantized data and those corresponding to next expansion layer of main coded layer are coded and layer coding steps are sequentially executed for all expansion layers to generate binary digit streams; additional information and quantized data corresponding to layer to be coded are represented as bits of same predetermined number and then they are arithmetically coded using predetermined probability model starting from sequences of more significant bits and to those of less significant bits, bit-modular data of left-hand channel and those of right-hand channel being alternately coded in terms of definite vectors. EFFECT: enhanced coding efficiency. 26 cl, 5 dwg, 43 tbl
