【作者】 王剑；

【导师】 林福宗；

【作者基本信息】 清华大学 ， 计算机科学与技术， 2004， 硕士

【摘要】 随着知识经济和网络时代的到来， 数字媒体产品的应用取得了惊人的进步。作为数字媒体产品知识产权宣告及保护的有效工具，数字水印技术自1993年第一次提出以来已经引起了人们极大的关注。特别是声音信号的数字水印技术已经成为近年来研究的热点之一。该技术就是向载体数据（如声音信号）中嵌入秘密信息以达到版权宣告及保护的目的。数字水印技术的关键就是工作域和嵌入策略的选取，从某种程度上说，工作域和嵌入策略选取的好坏从根本上决定了整个数字水印系统的优劣。而刚刚发展起来的小波变换是一种新型的时频分析方法，具有很多良好的性质，特别适合于声音信号的处理。本文正是将小波变换应用到声音信号的数字水印系统中，提出了两种较为理想的水印算法：基于小波变换域的数字声音水印和基于支持向量机的数字声音水印算法。通过各种常见的信号处理、压缩和解压缩及多种攻击性试验验证了这两种算法所生成的水印的性能。本文提出的基于小波变换域的数字声音水印算法，主要利用小波变换的多分辨率分析特性及通过分析人耳的听觉特性，选取合适的小波分解系数作为水印的嵌入载体。通过比较待嵌入水印值（二值序列）与对应嵌入位置的小波系数模2的值是否相等，来修改原系数值，以达到嵌入水印的目的。为增强嵌入水印的保密性，在水印嵌入前对水印进行了置乱和伪随机处理，不知道口令的人是无法正确恢复出水印的。 基于支持向量机(support vector machine，SVM)的数字声音水印算法，是把统计学习理论和机器学习的方法应用于数字水印技术中，利用支持向量机具有学习和自适应的特性实现了数字水印的半盲检测。其主要思想是在宿主声音中嵌入一段模板信息，定义模板信息与宿主声音之间的一种对应关系，将水印的检测问题转化为一个可用SVM处理的二分类问题, 利用SVM对先验知识（对应关系）的学习，以达到对未知数字声音水印的正确分类检测。更多还原

【Abstract】 With the economy of knowledge and era of Internet approach, the application of digital media works has made egregious progress. As a useful tool for the copyright protection and judgment, digital watermarking technique has gained more and more concerns in many ways since it appeared in 1993. Especially audio digital watermarking technique has been one of the research hotspots in recent years. With such technique we can embed secret information into digital audio signal, so as to arrive at the purpose of copyright protection and judgment. The chose of the work field and the embed method becomes more necessary, Wavelet transform, as a new powerful tool of time-frequency analysis, provides several good characters that make it appropriate to audio signal watermarking . In our work, two practical watermarking algorithms based on former studies are proposed: one based on wavelet transform and the other based on SVM (support vector machine). At the same time, we implement some experiments to test robustness and inapperceivity of the watermark produced from the two algorithm. The experimental results show that the embedded watermark is robust to audio signal processing. such as MPEG audio coding, cropping, filtering, resampling and requantizating.The algorithm based on wavelet transform makes good use of the characteristic of multi-distinguish and the sense of hearing characteristic to embed and extract the watermark. At the end, lots of simulated results are given to prove the performance and potential application of our system compared with other relevant systems.A novel digital audio watermarking technique based on SVM(Support Vector Machine) is proposed in this paper. The main idea of the method is that the retrieval of embedded watermark can be considered as a two-class problem, and SVM can be used to learn the characteristics of the embedded watermark in audio. Due to the SVM possessing the learning and adaptive capabilities, It almost exactly recovers the watermark from the watermarked audio.更多还原