Abstract: Digital watermarking techniques have been widely studied and applied as a main method for copyright protection.This paper proposes a framework that analyses each audio frame and selects positions for embedding adaptively using a psychoacoustic model.A new method for embedding digital watermarks into audio signals in the frequency domain based on the critical sub-band energy quantization is introduced.This framework can control the trade-off between the perceptual tran-sparence and robustness against MP3 attacks more efficiently.Experimental results show that the algorithm is robust to the most popular attacks such as MP3 compression,etc.