中文摘要：

针对一些水印算法存在的不足提出一种基于离散小波变换的适于版权保护的数字音频盲水印算法。算法对能量的概念作了新的定义，对其设计原理和过程作了理论分析和实验验证。算法实现时先分帧音频并对各帧作n级离散小波变换，然后计算n级小波系数近似分量前后两部分的能量比值，并放大；再将放大后的比值修改为奇数或者偶数以嵌入水印，最后按与嵌入时相同的比例缩小第n级小波系数近似分量的前半部分并作小波逆变换，得到水印音频。算法根据”级小波系数近似分量前后两部分能量放大后比值的奇偶性提取水印，实现水印的盲检测，提取水印不需原始音频。采用混沌加密水印信号，缩放因子也是密钥，算法安全性得到双重保护。仿真结果证明了算法的鲁棒性和不可感知性。

英文摘要：

The digital audio watermarking algorithm based on DWT （Discrete Wavelet Transform） fit for copy- right protection was proposed aiming to overcome shortcomings of some watermarking algorithms. The energy conception was redefined. Its design principle and process were analyzed theoretically and verified through ex- periments. In the implementation process, the audio signal was divided into frames and n-levels DWT was performed to those frames. The energy ratio of the first half and second half of the approximate component of the nth level wavelet coefficient was calculated and amplified by multiplying a constant coefficient and then modified to become even and odd according to the watermarking signal. The second half part of the approximate component of the nth level wavelet coefficient was kept unchanged whereas the modifird ratio was reduced and the first half of the approximate component of the nth level wavelet coefficient was transformed in correspondence to gain the watermarket audio signal. Watermarket bits were extracted according to the first half and second half magnified energy ratio parity of the approximate component of the nth level wavelet coefficient. Blind watermark detection was realized without need for the original audio and watermarks. Simulation results prove that the proposed algorithm is imperceptible and of good robustness..