Study On Digital Audio Watermarking Based On Content

With the rapid development of computer communication and network technique, the transmission, process and store of multimedia data become more and more convenient and quick. Simultaneously, it is easy to illegally copy, distribute and tamper. So the owner copyright of digital media should be protected, and the security and creditability of the multimedia data are threatened. Therefore, digital watermarking used as the technique of copyright protection and reliability authentication has been widely recognized by many researchers.This dissertation mainly focuses on designing the audio watermarking schemes. The fragile audio watermarking for data integrality authentication and robust audio watermarking algorithms for copyright protecting are proposed, respectively, The achieved innovative results are given as follows.1. The strength of watermarking is calculated adaptively by using SNR, and the right watermarking strength is gained by experiments over and over in existing algorithms can be avoided. On the condition of setting SNR, the maximum of the watermarking strength is obtained and the proposed algorithm can detect the watermark blindly. The dither quantization is used in embedding and extracting process and drawback of classical quantization method can be avoided. The dithering vector is generated by the chaotic system, and white noise property of chaos enhances the security of the algorithm.2. A fragile audio watermarking algorithm in NDCT domain content-based is proposed. The two quantization steps are deduced theoretically, the lackness that the appropriate quantization step is confirmed by multiple replication experiments in existing algorithms is avoided; the watermarking is generated by audio content and can avoid security problem of binary image watermarking in transmission; meanwhile, NDCT matrix elements are produced by linear function of chaotic sequences and it enhances system security furthermore. While authenticating audio integrity, the algorithm can also locate malicious tamper location.3. A robust audio watermarking algorithm based on ratio is designed. The watermark is embedded or extracted by quantizing the ratio. Today, many robust audio watermarking algorithms are coterminous with audio type. Because the proposed ratio-based algorithm is based on audio content, the robustness has nothing with audio type. The experimental results show the strong robustness.4. Synchronization attacks process of TSM, dithering, cropping and so on, and shortcomings that existed anti-synchronization attack audio watermarking methods are analyzed, and a new scheme based on audio global feature parameter and even dividing frames is proposed. By even dividing frames, synchronization place continuous offset can be avoided; the watermark are generated, embedded and extracted by stability of the audio global feature parameter. The mean in time domain is redefined, normalized center of gravity is defined and their stability for TSM manipulation is proved in theory and experiment. The stability of mean of wavelet approach coefficients has been proved by existed literature. The stability of the mean in the time domain, the mean of wavelet approach coefficients and normalized center of gravity for synchronization attacks is used for designing3algorithms against synchronization attack. Experimental results demonstrate the proposed scheme is effective. Compared with many algorithms that can only resist single type synchronization attack in existing literatures, the proposed watermarking algorithm can resist multi-type synchronization attacks.