Study On The Digital Audio Watermarking Scheme Based On Blind Source Separation

As the development and application of computers and Internet, multimedia data has become an important way for information acquirement. How to carry out copyright protection and ensure the security of the information has become one of the urgent problems to be solved in digital communication. As an effective way to solve this problem, digital watermark technique has attracted many attentions. At a word, digital watermarking is a scheme to insert some information that can represent the author's identity into the original data, to verify the property or the integrity of them.In this paper, three kinds of audio watermark techniques based on Blind Source Separation (BSS) are discussed. Firstly, the meaning of the task, the research background, and 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 BSS based digital audio watermarking methods are proposed. The idea of code mapping and Independent Component Analysis (ICA) are utilized to realize a time domain multi-bit robust watermarking scheme in the first method. Aim at the requirement for the application that need to protect the copyright of the multimedia and authenticate its safety simultaneously, a wavelet domain multipurpose watermarking scheme based on BSS is given in the second method. The embedding order of different watermarks is no longer an issue to be considered. The robust watermark is multi-bit and the fragile watermark can give the authentication result as well as the tampered areas. The third method, a dual-domain watermarking scheme based on BSS, gives a solution to overcome the sensitivity problem of time-scale attack for most existed audio watermarking schemes. It is robust against both common signal processing operations and time-scale attacks.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 watermarks, the result is intuitionistic; what's more, the fragile watermark in the second method can localize the tampered areas precisely.