基于奇异值分解的DCT域音频数字水印算法

Digital Audio Watermarking Algorithm Based on Singular Value Decomposition in DCT Domain

提出了一种基于DCT直流分量的音频数字水印算法。算法根据音频信号在DCT域能量较大处可以隐藏较多信息的水印嵌入原则,结合音频信号在DCT域能量分布特征,首先把原音频信号先进行分段处理,然后对每段做DCT变换,找出水印嵌入点,实现了在一段音频信号中隐藏一幅二值图像的奇异值向量。实验结果表明,该算法对MP3压缩、低通滤波、重采样等恶性攻击具有较强的鲁棒性和不可感知性。

As a main method for copyright protection, digital watermark techniques have been widely studied and applied. A new algorithm of digital audio watermarking based on DC components in DCT（Discrete Cosine Transform） domain is presented. According to the watermarking embedded principle that the place of large energy in DCT domain can hide more messages, combined with the characteristic of audio signal in DCT domain, the source audio signal is divided into segments, then the points that will embedded watermark signal on the segments＇ DCT domain are found. The algorithm realizes to hide a binary image in an audio signal. The analysis of watermark embed depth test and StirMark attacks are also presented. Studies have shown that the algorithm of digital audio watermarking has best result on robust and imperceptible, such as low pass filter, resample, MP3 compression and so on.