Abstract: In order to protect copyright, a novel cepstrum transform watermarking algorithm in wavelet domain was proposed. The proposed algorithm adopted the idea of Least Significant Bit (LSB) and quantization. Firstly, the audio signal was divided into several segments; the segment was transformed with wavelet domain and complex cepstrum. Then the LSB of mean of complex cepstrum coefficients was computed. Lastly, the watermark was embedded into the mean of complex cepstrum coefficients through changing its odd and even. When extracting, only the LSB of mean of complex cepstrum coefficients was judged to restore the watermark. The algorithm was blind. The embedding processing of the proposed algorithm does not depend on any threshold, and it is more conducive to the practical application of the algorithm. The experimental results show that the algorithm is very robust to re-sampling, re-quantization, low filter, amplitude attack, etc.

Key words: audio watermarking, Least Significant Bit (LSB), quantization, cepstrum

摘要： 为了达到版权保护目的，采用基于最不重要位（LSB）和量化思想，提出了一种新颖的小波域内倒谱水印算法。首先将音频数据进行分段处理和小波变换，并计算每段音频小波系数的倒谱域均值，然后计算每段均值的LSB位，最后通过LSB位的奇偶量化来完成秘密水印的嵌入。提取时，只需要判断倒谱域均值的LSB奇偶来恢复秘密水印。算法完全属于盲提取。算法嵌入过程中不依赖于任何阈值，更有利于算法的实际应用。实验结果表明，算法对于重采样、重量化、低通滤波和幅度攻击等都有很好的鲁棒性。

关键词: 音频水印, 最不重要位, 量化, 倒谱