Study On Digital Audio Watermarking Technique In Transformed Domain

As the development of computers and Internet, the application of multimedia data has become widely. Since digital product is easy to copy and revise, how to effectively protect copyright and legal right has become an urgent problem. Digital watermark technique becomes a new way to solve this problem. It mostly includes static image watermark, video watermark and audio watermark. At present, the study mainly focuses on the first two watermarks and the technique is comparatively mature. As digital audio compression technique is more and more mature, music products like mp3 is widely spread on the internet. Digital audio watermark embeds image which is on behalf of the authors' information, signature and ID of products into music products to fulfill copyright protection.In this paper, three kinds of audio watermark techniques are discussed. Firstly, the meaning of the task, the research background, the related concepts and characteristics are introduced. Some typical methods are reviewed as well. Secondly, basic background knowledge related is introduced. After that, three kinds of digital audio watermarking methods in transformed domain are proposed. Aim at the problems that most existing digital audio watermark schemes can weakly resist time-scale attack, a digital audio watermark scheme based on feature extraction and subsampling is proposed. Independent Component Analysis(ICA) is utilized to employ feature extraction which can strengthen the security of the scheme. According to the stabilization of Singular Value Decomposition(SVD), the second method, an audio watermarking scheme based on block SVD and dither quantization is proposed. The third method, a watermarking scheme based on DWT and DCT, satisfies the requirement for the application that need to protect the copyright of the multimedia and authenticate its safety simultaneously. It gives a solution to overcome the sensitivity problem of time-scale attack for most existed audio watermarking schemes; what's more, the embedding order of different watermarks is no longer an issue to be considered.Experimental results show that, all of the three methods can satisfy the requirements of transparency, robustness and security; the watermark signal can be extracted blindly; the robust watermarks are all multi-bit watermark and the result is intuitionistic; what's more, all of the methods can achieve self-synchronization.