自适应量化嵌入数字水印的音频隐密通信方法

A digital watermarking embedding scheme of self-adaptive quantization-based for audio hidden transmission

本文提出了一种基于数字水印的音频隐秘通信方法。鉴于隐藏效果的直观性,原始水印信息以二值图像为例,也可为其他类型如语音。该算法首先对有具体含义的原始水印信息进行预处理,使之成为适于嵌入音频的一维数字水印序列;然后对音频载体分段,根据各段能量大小选定隐藏段,再对各段进行离散余弦变换(DCT);考虑到不可见性和鲁棒性的折中,本文依据噪声模型确定的噪声敏感度因子在DCT域的中低频系数区选定具体嵌入位置;并采用量化嵌入方法,可实现盲提取。实验结果表明该方法具有良好的不可见性和鲁棒性,能抵抗滤波、重采样、重量化、加噪声等常规信号处理攻击。

This work proposes an approach based on digital watermarking for audio hiding transmission. To achieve human intuition, takes binary image as original marking information, or audio. The algorithm pre-process the meaningful original marking information to obtain digital marking sequences to embed into the audio, segmenting the audio signal into regions; determining the hiding regions according to the signal energy; applying discrete cosine transform （DCT） to the regions. With consideration of good robustness and hidden nature, choosing specific embedding regions according to noise sensitivity factor corresponded in low frequency coefficient of DCT region in a noise model, implement blind extracting with quantization embedded method. The experimental result shows that the embedded data are robust and blind, strong capacity of resisting common signal processing attack, such as filtering, re-sampling, requantization, adding Gaussian noise corruption.