The Research Of Digital Audio Watermarking Algorithm Based On The Transform Domain

With the rapid development of networking techniques and the maturity of the multimedia compress technique, the multimedia technique based on the network has obtained the widespread application, such as video frequency, audio frequency, image and digitized product. So how to protect the copyrights of these digital products in the network environment has become a pressing problem. Digital watermarking technology is put forward in such a context. Hiding some secret information is a way to protect the data security of digital products. Digital watermarking technology can effectively solve the protection of the copyrights problems of the digital products on the network. It has obtained a wide range of research and application, and has become a hot research at home and abroad.Audio watermarking technology is the watermarking technology based on the digital audio protection. Through hiding secret watermarking or some information in the audio files to protect the copyrights, this paper studies the audio watermarking technology based on the transform domain. Combining the DWT and DCT, it proposes two audio watermarking algorithms based on the transform domain. It introduces the research background of digital watermarking, general principles, application and the status quo. It focuses on the definition of audio watermarking, characteristics, basic model, evaluation criteria and the application. For existing audio watermarking algorithms, it provides a simple summary and classification. The main research contents and results achieved are as follow.1. Propose an audio blind watermarking algorithm based on DWT and coefficients comparison. It is a new algorithm through quoting literature[25] and literature adaptation[19], literature proposition[22]. Literature[19] proposes a non-blind audio watermarking algorithm based on DWT. This approach requires the original audio in time of detection. It is not a blind detection. Therefore, it has so many restrictions in practical applications, it is useless. Literature[22] proposes a blind audio watermarking algorithm based on DWT and local minimization value. Its basic idea is to search the"local extreme point"in the low-frequency coefficients of DWT, by modifying the local minimum and the magnitude of the value of adjacent points to achieve the watermarking embedding. The disadvantages are, first, it is difficult to find the local extreme value of the low-frequency coefficients. Second, it is very difficult to realize and compute. Literature[25] proposes an image watermarking algorithm based on DWT and transform domain coefficient comparison. Although these algorithms have a certain degree of robustness and transparency, those have some shortcomings more or less. So this algorithm considers the transparency which is an important factor for the watermarking algorithm. It aims to improve the algorithm's simplicity and transparency in the premise of ensuring the algorithm's strong robustness, takes fully advantage of wavelet transform fast, simple and multi-resolution features, uses a binary image as the watermarking signal, determines the watermark's embedding position according to the watermark themselves, makes use of the SNR to determine the embedding strength, embeds the watermark signal into the low-frequency after the audio's wavelet transform. The advantages of this algorithm are as follows.(1) Algorithm is simple and easy to achieve, and its anti-jamming ability.(2) Using the Arnold on the watermark image scrambling to ensure the security of the watermarking image.(3) The embedding process is first to DWT the original audio, the embedding location selected is the largest of its energy-low-frequency part of the audio. It is very useful to cover up the impact of the watermarking, and the algorithm has high robustness.(4) The watermark has a dual role. First, it is used to determine the embedding location of each segment by a continuous three watermarking signal; second, the coefficients values of the embedding location, the adjacent location and the watermarking signal can decide whether to modify the value of the embedding position.(5) When detecting, as long as there is a certain degree of readability of the audio, we can detect the watermarking image.(6) The adaptive signal to noise ratio can determine the strength of the watermarking embedding in each segment and it increases the robustness of the algorithm.2. Propose a blind audio watermarking algorithm based on DWT and DCT mixed-domain. The basic idea of literature[28] is, first, uses the DWT to transform the whole original audio, divides them into several bands, according to the frequency of human auditory masking effect, selects the low-frequency coefficients to be the embedding locations, carries out them into stages. Then uses the DCT to transform the low-frequency coefficients, selects the coefficients of the DC, makes the DC coefficients to be zero, last according to the value of the watermark bit to determine its value. When the watermark bit is'1', sets the DC coefficients to beΔ, when the watermark bit is'0', sets the DC coefficients to be–Δ, theΔis the watermark embedding strength. Literature [32] proposes an effective digital audio watermarking based on the mixed-domain. Its basic idea is that using the DWT to transform the audio section, and then selects the high and low frequency coefficients, transforms the two kinds of coefficients by the DCT respectively to embed the watermarks. Such programs have certain advantages in the anti-simultaneous attacks; literature[33] proposes a new transform domain-based digital audio watermarking algorithm. It is first to transform the original audio by 5 DWT, and then selects the five decomposition under the low-frequency coefficients, transforms the coefficients by DCT, at last selects the lager value of the DCT coefficients to embed the watermark. This algorithm has good transparency, it has a good robustness when meeting the noise, heavy sampling, weight, and so on attacks. The algorithm proposed in this chapter is different from the traditional audio watermarking based on the mixed domain. It can make up the weakness of the traditional mixed domain audio watermarking algorithms'good robustness but low transparency and high transparency but low robustness, introduce the concept of the"zero watermark", make a full use of the characteristics of DWT and DCT transform, using the strong robustness when embedding the watermark in the low-frequency coefficients, according to the DC component's positive and negative nature of the data is not easy to change, the DC components have greater feeling of capacity characteristics than AC components and"zero watermarking"algorithm's high transparency. The basic idea is: using a photo as watermarking. First, to transform the audio by three DWT as a whole, select 3 decomposition under the low-frequency coefficients, and then sub them for each segment to DCT, select the first factor of each segment-DC component, then , according to the positive and negative DC component , construct a 0-1 sequence of the"zero watermarking". When the DC component is greater or equal to 0, we extract 1, or else extract 0, use the series of extracted watermarking information with the embedded watermarking for XOR, so we construct a unique"zero watermarking"sequences for copyright protection. This algorithm is to make up the shortcomings of the literature [32] such as low transparency, literature [32]'s weak robustness, literature[32]'s computational complexity, we reach a good balance among the transparency, robustness and the complexity.Chapter 3 presents an audio watermarking algorithm based on traditional DWT, which fully takes into account the transparency without affecting the robustness; chapter 4 proposes a new audio watermarking algorithm which is very different from the traditional algorithms. It is based on the"zero watermarking"concept. The watermarking image doesn't really embed into the audio carrier, so its transparency is so high. When extracting the watermarking, we can clearly extract the watermarking signal. What's more, they don't need the original audio participation, and it is blind detection. The experiment results show that those two are very effective audio watermarking algorithm, and they can be widely used to protect the copyrights of the digital audio products.