Study On Robust Audio Watermarking Algorithms In Transform Domain

With the rapid development of internet and audio compression techniques, digital audio signals become much easier to transmit, at the same time, the need to protect audio works has become more and more important. The digital audio watermarking technique is a kind of effective method to protect audio copyright, hiding some secret information is a way to protect the integrity of audio works and author's right. The author mainly research robust audio watermarking technique based on the transform domain in this paper.Firstly, the research significance and the basic theory of audio watermarking techniques, and a summary of current research at home and abroad are introduced in this paper. Then combining the DWT(Discrete Wavelet Transform) and EMD(Empirical Mode Decomposition), the author designs a robust audio watermarking algorithm, which could resist all kinds of common signal processing operations. Finally, the author proposes a new audio blind watermarking algorithm based on auditory analysis in wavelet domain.The main work includes the following respects:1) Based on instances at present, the author analyzes the characteristics of ear hearing system and the typical robust audio watermarking algorithm in time domain, frequency domain and compressed domain. These work lay a good fundation for my future research.2) The basic principle and characteristics of DWT and EMD are introduced. DWT has good time-frequency properties, and it can watch signals from coarse to fine gradually, EMD algorithm is based on the local features of audio signal, and the signal can be self-adaptive decomposition by EMD method. Baesd on the theory and experiment, the author find that the mean trend is stable for many attacks. So a blind audio watermarking algorithm is proposed, which based on multi-resolution analysis of wavelet transform and adaptive property of EMD method. The author apply EMD algorithm to decompose the low-middle frequency coefficients of DWT, because the mean trend is not easy to change, so it is selected to embed the watermark. This algorithm is robust to most attacks based on Cool Edit Pro 2.0 and Stirmark.3) An adaptive audio watermarking scheme based on human auditory masking effect and wavlet packet decomposition is proposed. Firstly, non-uniform subbands are obtained by wavelet packet decomposition, which are closed to human auditory system. Then the masking threshold and energy are calculated in the low-middle frequency coefficients of wavelet domain. The watermarks are embedded in appropriate subbands. Both the subbands and quantization step are adjusted adaptively based on masking threshold and energy. The experiment results show that the proposed scheme is inaudible and robust.