With the rapid development of computer multimedia technology and the Internet, the exchange of multimedia information has reached an unprecedented depth. Meanwhile, piracy and tampering have become rampant, so the digital watermarking technology emerged, and received extensive attention. This dissertation aims at digital audio zero-watermarking and dual watermarking schemes based on zero-watermarking. The works and efforts are described as follows:(1) Utilizing the nature of RBF network, simple structure, good nonlinear approximation ability and fast convergence, a digital audio zero-watermarking algorithm in wavelet domain was proposed based on RBF network. In our application, the relational model was established between the important characteristics of the original audio and the watermark by using the DWT coefficients as the input signal and the watermark as the training target. Our system constructs meaningful watermark—different from what others have done. The experimental results show that the scheme does not change the original audio's perceived quality, has excellent ability to reject common attacks, and achieves blind detection. Compared with the scheme completed by BP network under the same experimental environment, the scheme is faster when dealing with the watermark embedding and extraction, has more accurate outputs, and realizes easier.(2) Aiming at the problems of previous dual watermarking algorithm, poor robustness and inaccurate localization, a digital audio dual watermarking scheme based on wavelet packet analysis was presented. The scheme combined with the image segmentation can link the image block and the original audio closely, and then embed robust watermarking and semi-fragile watermarking by introduction of wavelet packet coefficients. The robust watermarking based on the concept of zero-watermarking allowed for copyright protection on the audio; the semi-fragile watermarking embedded watermark image block by mean quantization was able to ensure the authenticity of digital audio, where two-dimensional image did not need to be converted into one-dimensional sequence which was superior to previously proposed ones. In this article, since robust watermarking and semi-fragile watermarking were inserted into different wavelet packet coefficients, they did not affect each other. Accordingly, embedding or extraction of robust watermarking and semi-fragile watermarking can be implemented simultaneously while the order problem in embedding or extraction of multiple watermarking did not take into account. Simulation results demonstrate the robustness of zero-watermarking algorithm has been improved which can protect the copyright of audio better; Moreover, The semi-fragile watermarking has good robustness in resisting a series of conventional treatment, not only can detect and localize malicious attacks, but also can improve the accuracy of localization, so it efficiently achieves authenticity and content integrity certification of audio. |