【作者】 付丽娟；

【导师】 李平； 薛居宝；

【作者基本信息】 山东大学 ， 光学工程， 2008， 硕士

【摘要】 数字水印作为一种有效的数字产品版权保护的技术手段,是目前国际上信息安全领域内的一个前沿课题。随着计算机以及因特网的普及,音频数字产品得到日益广泛的普及和应用,有效提高了信息传播的效率和准确度。音频数字水印技术有效地防止了个人和团体,在没有作者的许可的情况下,对数字产品随意复制、篡改、销售等,相当程度上解决了数字产品产权保护的问题。数字音频水印的应用与其它媒体上的水印应用类似,主要有三个:版权保护、盗版跟踪和认证,其中版权保护是最迫切和重要的。与数字图像和视频水印技术相比,数字音频水印技术面临着更大的挑战。一方面是因为人类听觉系统(HAS)对随机噪声十分敏感,使可以嵌入的水印数据量非常有限;另一方面在互联网上可以自由得到众多的音频编辑工具对数字音频的结构进行修改,从而对水印的生存构成了巨大的威胁。文中介绍了数字水印的一些相关知识,分析了现有的一些音频数字水印技术的工作原理。在此基础上以直观的二值图像作为水印信息,以音频数据为嵌入对象,在详细分析数字音频水印技术特点和需求的基础上,对现有部分算法进行了分析并提出改进的方向。文中还探讨了;基于DCT域的自适应音频数字水印算法;和基于模数运算的DWT域数字音频水印算法。第一种算法是将二维二值水印图像降维后,与m序列做扩频调制,再对数字音频信号进行分段处理,并依据人类听觉系统(HAS)择段做离散余弦变换(DCT),在离散余弦变换域内,把调制过的水印信号经过量化处理,嵌入到离散余弦域系数的幅值上,以完成水印信息的自适应嵌入。第二种算法通过每帧前两节精细分量的能量比较,又由水印比特为“1”或“0”,采用不改变或缩小精细分量的方法来嵌入水印。提取水印时则对音频信号做相应的小波分解,从低频系数中提取{0,1}序列,这样便可以还原水印信号。实验证明两种方法都不改变原始信号的感知质量,算法简单计算量小,具有较强的鲁棒性和较好的透明性。两种算法都可实现盲检测。更多还原

【Abstract】 As an effective way of protecting the copyrights of digital products, Digital Watermark is the leading subject in information security fields internationally. As the popularization of computers and internet, digital audio products have been used more and more widely which dramatically improved the efficiency and precision of the transmission of information. Digital Watermark can effectively prevent individual or organization from copying, tampering or selling products without the permit of the authors, which significantly resolved the copyright problem of digital products.The use of Digital Audio Watermark is very similar to the use of Digital Watermark in other media, it mainly includes copyright protection, pirate trace and certification. Compared with Digital Image and Video Watermark, Digital Audio Watermark faces more challenges. On one hand, the Human Auditory System(HAS) is very sensitive to random noise,so the embeddable watermark is limited, and on the other hand, there are so many audio edit tools on the internet to change the structure of digital audios, which threatens the survival of watermark.This paper introduces the interrelated knowledge of Digital Watermark, analyzes the principles of some present audio watermark technologies. Using intuitive binary image as watermark information and audio data as the embedded object, based on detailed analysis of the features of audio watermark technologies and requirements, I analyzed some of the present algorithms and proposed directions of improvement.The self-adapting Digital Watermark algorithm in DCT domain and Digital Watermark in DWT domain based on modular arithmetic are also discussed in this paper. The former algorithm works like this: first reduce the dimension of the two-dimension binary image watermark, and do spread spectrum modulation with an m sequence, then cut the digital audio signals into sections and choose sections according to HAS to do the Discrete Cosine Transform (DCT), in DCT domain, quantify the modulated watermark signals and then embed them in DCT domain or amplitude to finish the imbed of the self-adapting watermark information. The latter algorithm compares the energy of the first two delicate components of each frame, according to the watermark bit "1" or "0", by unchanging or reducing the delicate components to embed watermark. When extracting watermark, the audio signals will be decomposed with relevant wavelet, and extract {0,1} sequences from low frequency factors to restore watermark signals. Experiments prove that neither of the two algorithms changed the perceptual quality of the original signals and both of them had the advantage of simplicity, small amount of calculation and good robustness and transparency. Both the algorithms can apply blind detection.更多还原