The invention discloses a method and device for spatial parameter stereo coding/decoding. Firstly, the coding end down mixes the signals of the left and right sound tracks in the time domain or frequency domain; then the down-mixed signals are transmitted to a transition coder to produce monophonic coding data; at the same time, down-mixed signals containing quantization errors are extracted, and the transform domain of the signals is divided into continuous non-overlapping sub-tapes according to the instantaneous spectrum, and he sub-tape is used as a unit to extract the spatial parameters of the left and right sound tracks. The decoding end generates down-mixed signals identical with the coding end and containing quantization errors according to transform coding data and achieves division of sub-tapes according to the same dividing method; then the sub-tape signals of the left and right sound tracks (stereo) are rebuilt according to the spatial parameters, and time-domain stereo signals are exported by inverse transform. Compared with the prior art, the invention has no additional delay caused by positive and inverse tape transform and capable to realize time-frequency division under the condition that sub-tape division information is not transmitted. And as a result, the real-time property and efficiency of the spatial parameter coding is improved.