Abstract This paper proposes a robust reversible audio watermarking scheme by introducing a high-order difference statistical model. In the proposed scheme, the high-order difference statistical model is constructed to compute a statistical histogram. The histogram can be regarded as a robust feature and shifted for embedding a watermark sequence. Firstly, an original audio file is divided into a number of frames, then a high-order difference statistic in a frame can be calculated to construct a histogram of the high-order difference statistics. With a hiding key, the watermark can be embedded into the audio file by shifting the histogram. In the process of watermark extraction, with the hiding key, the watermark can be correctly extracted from the shifted histogram. Furthermore, by using the inverse operation of histogram shifting, the original audio file can be restored non-destructively. In addition, the watermark can still be correctly extracted from the watermarked audio file that is attacked by some signal processing operations to some extent, such as MP3 compression of 48 Kbps and Gaussian noise of 25 dB. Compared with the existing several robust reversible audio watermarking schemes, the proposed scheme has better performance in audio quality and robustness. Experimental results have shown the effectiveness of the proposed scheme.

中文翻译：

---

基于高阶差分统计的鲁棒可逆音频水印

摘要 本文通过引入高阶差分统计模型，提出了一种鲁棒的可逆音频水印方案。在所提出的方案中，构建高阶差分统计模型来计算统计直方图。直方图可以被认为是一个鲁棒的特征，并且为了嵌入一个水印序列而移动。首先将原始音频文件分为若干帧，然后计算一帧内的高阶差分统计量，构建高阶差分统计量的直方图。使用隐藏键，可以通过移动直方图将水印嵌入到音频文件中。在水印提取过程中，利用隐藏密钥，可以从平移后的直方图中正确提取水印。此外，通过使用直方图移位的逆运算，可以无损地恢复原始音频文件。此外，在受到一定程度的信号处理操作攻击的带水印音频文件中，如48Kbps的MP3压缩和25dB的高斯噪声，仍然可以正确提取水印。与现有的几种鲁棒可逆音频水印方案相比，所提出的方案在音频质量和鲁棒性方面具有更好的性能。实验结果表明了所提出方案的有效性。与现有的几种鲁棒可逆音频水印方案相比，所提出的方案在音频质量和鲁棒性方面具有更好的性能。实验结果表明了所提出方案的有效性。与现有的几种鲁棒可逆音频水印方案相比，所提出的方案在音频质量和鲁棒性方面具有更好的性能。实验结果表明了所提出方案的有效性。