一种抗D/A和A/D变换的音频水印算法

音频水印的许多应用要求水印能在模拟环境中保留下来，这涉及到水印的D／A和A／D（以下简称DA／AD）变换，然而，音频水印抗DA／AD变换尚末引起人们的充分重视，已有的音频水印算法能抵抗DA／AD的很少，文中通过大量的实验，分析并得出了在DA／AL）过程中影响音频水印的主要因素：波形失真和时间轴线性伸缩，并针对这些因素提出了一个基于DWT的抗DA／AD变换的茸音频水印算法，针对音频水印DA／AD变换过程中时间轴上的线性伸缩和因幅值变化与噪声影响产生的波形失真，作者在算法中分别采取了相应的对策：（1）在检测过程中结合同步码蕈定位和线性伸缩恢复来消除时间轴上线性伸缩带来的影响；（2）针对幅值改变，在DWT域采用了基于三段低频小波系数之间能量关系的嵌人对策；（3）自适应调整嵌入强度来满足抗噪声攻击的要求，实验表明，所提出的音频水印算法具有很强的抗DA／AD攻击性能和抗其它各种通用的音频处理和攻击的能力。

An Audio Watermarking Algorithm Against the D/A and A/D Conversions

In many applications of audio watermarking, the watermark is required to survive well in analog environments, which involve the D/A and A/D conversions(denoted as DA/AD in this paper). However, the robustness of audio watermarking against the DA/AD processing has not drawn sufficient attention yet. According to our best knowledge, there has no report for the audio watermark to effectively resist the DA/AD conversions. Based on extensive experiments, this paper analyze and conclude the main degradations on the audio watermark caused by the DA/AD conversions, waveform distortion and temporal linear scaling. Accordingly, the authors propose a DWT-based blind audio watermarking algorithm against the DA/AD conversions. To resist waveform distortion of the watermarked audio, they adopt the embedding strategy using the relative energy relations among groups of the wavelet coefficients in low-frequency sub-band and adaptively control the embedding strength. The resynchronization is designed to resist temporal linear scaling. The experimental results show that this algorithm is robust to the DA/AD conversions, and other most common audio processing manipulations and attacks.