Abstract: This paper proposes a Discrete Wavelet Transform（DWT）-based robust and blind audio watermarking algorithm to resist cropping attack.The approach initially divides the audio signal into several frames including the same samples.Each frame is then transformed by DWT,aiming at extracting the low frequency coefficients of every frame and the corresponding absolute mean.According to comparison between the absolute mean and the low frequency coefficients,a binary image via Arnold pre-processing is embedded in audio low frequency coefficients in wavelet domain.Experimental results show the performance of the method is ideal under various signal processing manipulations such as Gaussian noise,resampling,MP3 compression,random cropping,jettering,especially,robust to cropping attacks.

摘要： 提出一种基于离散小波变换的有效抵抗随机剪切的鲁棒性音频水印算法。该方法首先把音频信号划分为若干个包含相同采样点的帧;然后对每帧进行小波变换,提取小波低频系数并计算相应的绝对平均值;最后根据每帧绝对平均值与低频系数比较的结果,将经过Arnold变换后的二值图像嵌入到小波低频系数中。该算法在对水印进行提取和检测时不需要原始音频。实验结果表明,提出的方法在抵抗各种音频处理（如高斯噪声、MP3压缩、重采样、去噪处理、随机剪切、抖动攻击等）性能良好,尤其是抵抗音频信号的恶意剪切攻击鲁棒性更强。