Research On Robust Audio Watermarking Algorithms

With the development of Internet and digital multimedia technologies, it's convenient for people, but it also easily lead to illegal copy and distribution, which against intellectual property. Digital music works are widely disseminated, so it's necessary to provide copyright protection for these works. Robust digital audio watermarking has become one of the research hotspots on copyrights potection in recent years.The article is commited to robust digital audio watermarking, and the following work has done:1,The existing robust watermarking algorithms based on psychoacoustic model I is not blind extraction,which need original signals, or through the complex correlation compulation to extract watermarking, and other issues. So a new robust digital audio watermarking algorithm based on psychoacoustic model I is brought. Frequency points for embedding watermarking are selected by the stable sort of chaotic sequence. And using the corresponding masking values of frequency points for embedding watermarking to regulate secret information, so compulation of the proposed algorithm is small, meanwhile, it's adapt to existing compression standard. And the process of extraction is simple.2,Because of the fixed frequency points, traditional disctete Fourier transform may cause potential safety problems when it's used in watermarking algorithm. Propose robust digital audio watermarking algorithm based on the nonuniform discrete Fourier transform(NDFT). This scheme is designed to compare magnitude coefficients of NDFT for watermarking's embedding, and it can resist on malicious attack, that a class of quantization watermarking algorithm can't achieve. At the same time, chaotic pseudo-random sequences generator based on nonlinear dynamic filter(NDF) security model are constructed to select the frequency hopping points for NDFT, improving the overall system security. 3,Aiming at replay attack of networks in the remote speaker recognition systems and distributed databases speaker recognition systems, a time-stamp watermarking method is brought, which quantizes magnitude coefficients of NDFT. The time-stamp watermarking is embedded into the user's voice sample/voice characteristic information to resist on the replay attack. Considering the practicality of the algorithm in network, chaotic encryption and error correction coding, chaos synchronization sequences search, chaos scrambling have been taken.