一种新的基于压缩感知的稀疏音频水印算法

A New Sparse Audio Watermarking Algorithm Based on Compressive Sensing

提出了一种基于压缩感知的音频水印算法.该算法利用数字水印和音频宿主信号在离散余弦变换域下都具有稀疏性,将水印经过压缩感知处理后,嵌入于音频载体信号的离散余弦变换系数中;在水印提取时,利用压缩感知的去噪原理和重构方法,在不利用原宿主信号情况下,将水印提取出来,此提取完全是盲提取.计算机仿真实验表明,这种算法能提高水印在高斯噪声攻击下的鲁棒性.

This paper presents an audio watermark algorithm based on compressive sensing （CS）. The algorithm takes the advantage of the feature that the digital watermark and host audio signal are both sparse in discrete cosine tran6form（DCT） domains. The watermark that has been processed embeds the host audio signal＇s DCT coefficients. This audio watermark extraction is done completely blindly by the CS denoising theory and reconstruction method. Computer simulation shows that this algorithm can increase watermarks＇ robustness under the attack of the gaussian noise.