基于重心的抗同步攻击的音频水印算法

Audio Watermarking Algorithm Against Synchronization Attacks Based on Gravity of Center

抗同步攻击是音频水印领域公开的难题.相对重心是音频的一个稳定的全局特征量,对TSM(Time scale modification)等同步攻击不敏感.本文提出一种基于相对重心的抗同步攻击的音频水印算法,提出的算法通过量化音频的相对重心生成水印,生成的水印由混沌序列加密后以量化的方法嵌入载体音频三级小波近似分量均值中.为了降低同步攻击后同步位置的连续偏移,水印嵌入提取过程采用了均匀分帧的方法.实验结果表明提出的算法能抵抗常规信号处理操作,同时能抵抗20%的TSM操作(包含音调不变TSM和重采样操作)和1/10的抖动操作,有很好的鲁棒性.

It is an open difficult problem in audio watermarking field to resist synchronization attacks. An audio watermarking algorithm against synchronization attacks is proposed based on the relative gravity of center, which is a globally steady feature parameter of audio and is insensitive to time scale modification （TSM） operation and other synchronization attacks. The watermark is generated by quantifying relative gravity of center in our method. Firstly, the generated watermark is encrypted by chaos sequence and then the encrypted watermark is embedded into the third level wavelet approach coefficients of audio by quantification method. The method of dividing frame evenly is applied to embedding and extracting watermarking in this paper in order to reduce synchronization place offset continuously after synchronization attacks. Experiment results demonstrate that the proposed algorithm can resist common signal processing and 20 ~ TSM （including pitch-invariant TSM and resampling manipulation）, 1/10 dithering manipulation and has very good robustness.