Abstract: In order to protect the copyright and guarantee the hidden communication better,the model which can get robust and fragile watermarks with the same embedded watermark through the different keys is established.And then the blind extracting audio discrete watermarking algorithm based on polarity quantification of uniform quantitative the high frequency of total wavelet energy is proposed.The new algorithm takes the characteristic value of the watermark as the key of watermark pretreatment.Experimental result shows that the project has better robustness and fragility.