用于版权和内容认证的音频水印

提出了一种多功能水印算法，能同时对数字音频版权和内容完整性进行认证。首先将代表版权信息的二值图像以零水印方式嵌入音频美尔倒谱系数（MFCC），再将该系数转换成二进制串，作为最终的水印信息以量化方式嵌入音频离散余弦变换（DCT）域。检测时可实现两种水印信息的盲提取。实验结果表明，该方法不但对音频内容的恶意篡改非常敏感，而且能够定位出被篡改音频的位置；此外，还能有效抵抗诸如滤波、MP3压缩等常规信号处理操作，正确恢复版权信息。

Audio watermarking for copyright verification and content authentication

A multipurpose watermarking scheme was proposed to provide a solution for some applications that require both copyright protection and content authentication. A robust watermark, in the form of a binary image, was first embedded in Mel-Frequency Cepstral Coefficient (MFCC). Then, the watermarked MFCC was coded to be binary string as a fragile watermark to be embedded in the DiscreteCosineTransform （DCT） domain. Blind detection was achieved for both robust and semi-fragile watermarks. The experimental results show that the proposed method is sensitive enough to detect malicious modifications and is capable of locating the regions that have undergone malicious manipulations. Besides, it is also effective to detection reliability, embedding capacity and robustness against common signal processing operations.