Digital Audio Watermarking Based On Transform Domain And SVD

With the development of Internet technology and the emergence of audio formatssuch as MP3, the spread of the digital audio media become easier through the net workand its transmission became more efficient and effective. This makes it possible thatmalicious people infringe and tamper on the products of the digital audio.In order tosolve the increasingly severe copyright issues, digital audio watermarking technologyemerged at the right moment. Right now, it has become one of the hot topics of thecurrent academic researches.Although a great number of audio watermarking methods have emerged untilnow, they can not practice for real time applications. Their robustness is not strong.Besides this, all of the algorithm have the same shortcoming that their capacity is low.The singular value decomposition method is simple and can improve therobustness of the algorithm due to its stability. Considering the practicality of thealgorithm, two robust blind audio watermarking algorithms which have a certainpracticality are proposed based on the theory of the transform domain andsingular value decomposition. These algorithms have strong robustness, meanwhilethey all increase the capacity.The audio watermarking method based on mixed transform domain and SVD isstudied. The host audio is cut into pieces and each of the pieces has two parts. Thesynchronization code is embedded in the first part, while the watermark is embeded inthe second part. DWT is performed on the latter part. And then, DCT is performed onthe approximate component. After that, SVD is performed on the low-frequencycoefficients. Finally, the information of the watermark is embedded in the singularvalues. Simulation results show that the algorithm has achieved the compromiseamong imperceptibility, robustness and capacity.Another audio watermarking algorithm based on discrete wavelet domain andSVD is also built. In order to improve the security of the algorithm, the watermarkimage is scrambled by the Arnold transform before the watermark is embedded. TheDWT is then performed on the segments which are embedded the wartermark. Theapproximate component is selected to calculate the absolute mean value of every tenDWT coefficients. Finally, perform SVD on the matrix which is made of mean valuesand embed the watermark information into the singular values by quantization method. Experiments results show that this algorithm has good imperceptibility, strongrobustness and a higher capacity compared to other methods.