一种基于音频内容的NDCT域脆弱音频水印算法

Content-based Fragile Audio Watermarking Algorithm in NDCT Domain

针对音频信号内容完整性鉴定的需要,提出基于音频内容的非均匀离散余弦变换的脆弱音频水印算法。算法中水印通过量化音频三级小波近似分量的平均值而产生,并对三级小波近似系数做非均匀离散余弦变换NDCT,将生成的水印通过量化的方法嵌入到NDCT域第2个系数(即首个交流系数)上。非均匀离散余弦变换的频率采样点由密钥控制的混沌产生,提高了水印系统的安全。仿真实验表明:提出的算法对MP3有损压缩、滤波、重采样、量化等信号处理操作十分敏感,对恶意的替换操作等非同步攻击能够报警并能做出篡改定位。

In view of the need of audio content integrity authentication, the content-based non-uniform DCT fragile audio watermarking algorithm was put forward. The watermark was generated by quantifying the third-level audio wavelet approximation coefficients. The watermark was embedded into the second coefficient （namely the first active coefficient） in the non-uniform DCT （NDCT） domain. The embedding method was to quantify the active coefficient. The NDCT frequency sampling points were set by key-controlled chaos and security of the watermark system was improved. Simulation experiments demonstrate as follows： The proposed algorithm is extremely sensitive to signal processing manipulations such as MP3 lossy compression, filtering, re-sampling, quantization,etc; the algorithm can detect and localize illegal tampering.