In this paper, we propose a non-blind audiowatermarking scheme based on compression-expansion of stereo signals. We also demonstrate the general methods for embedding binary code in single channel signals and stereo signals even though there is only one possible state change. The proposed audio-watermarking method makes use of compression-expansion of adjacent frames of samples of the audio signal. Some frames of the signal are thus ‘distorted’. The distorted frames can be readily identified by comparing them with the original signal. The distorted frames can be used to embed hidden information. By applying psychoacoustic model in the compression-expansion process, it is found that the distortion is hardly discernable. The performance of the method under several types of attacks is assessed. It is found that when stereo signals are used, higher percentage of recovery can be achieved than when single channel signals.