一种新的基于量化的音频数字水印算法

A new digital audio watermark algorithm based on quantification

文章提出了一种新的基于量化的音频数字水印算法 ,该算法首先将视觉上可辨的二值水印图像降为一维序列 ,再将水印序列伪随机排序 ,最后对数字音频信号进行分段 DCT(离散余弦变换 ) ,水印信号通过量化处理过程嵌入到 DCT变换后的系数中 ,提取水印信号无需使用原始数字音频信号。仿真实验表明水印是不可察觉的 ,经过 mp3有损压缩、低通滤波、重采样等操作后仍具有很强的稳健性。

This paper presents a new digital audio watermarking algorithm based on quantification in DCT domain. First, the visually recognizable binary watermark image should be converted into the onedimensional sequence in the algorithm.The sequence is permuted pseudorandomly as the watermark. Then the digital audio signal is converted into segments and discrete cosine transform is performed. Finally the watermark is embedded in coefficients in DCT domain by quantifying process. The method can extract the watermark without the original digital audio signal. The experimental results show that the watermark is imperceptible and the algorithm is robust to digital audio signal, such as mp3 loss compression, low pass filter, resample and so on.