Research On Audio Digital Watermark Algorithms

Recently, with the rapid development of digital techniques, multimedia products such as image, video, audio, have been popularized and applied increasingly and widely, the efficiency and correction of communication have been improved clearly. But the side effect with it is that, some persons and organizations can easily copy, edit and transmit the content without the permission of the copyright ownership. So, how to take measures to enable copyright protection and information safety in the network, has become a problem that should be solved immediately. Digital watermarking technique is the most effective method to solve copyright ownership. But the current watermark algorithms applied widely are related with image and video. With the popularity of the new compression model mp3, more importance must be put on the copyright of audio. This dissertation selects two-dimension binary image as watermark information, and selects audio as embedding object, and embed the watermark in the coefficient of the DFT domain, the DCT domain, the DWT domain. The main research work and contributions of the dissertation are as follows:(1) This dissertation reviews the relative issues and current research status of audio digital watermark.(2) In the discrete Fourier domain, the watermark information is embedded by quantifying the amplitude of the DFT domain coefficient; hi the same time, the quantification procession and the provision that must be met in the modifying of amplitude is given. The watermark can be extracted without the original audio. The performance of algorithm can be unproved applying spectrum-spread.(3) hi the discrete cosine domain, the watermark is embedded by quantifying the coefficient of the DCT domain and the quantification procession is given; the watermark can be extracted without the original audio. Applying spectrum-spread, the adaptation algorithm and the blind algorithm are discussed, the differencebetween them is given.(4) In the discrete wavelet transform domain, the watermark sequence is modulated with m-sequence at first, then the watermark information is embedded by quantifying the coefficient of the transform domain. The quantification procession is given. The watermark can be extracted without the original audio. The performance of the watermark algorithm can be improved by applying the adaptation algorithm and embedding the watermark information in the largest coefficient of near part of the wavelet decomposition.(5)The performance of the watermark algorithm can be improved by combining the discrete wavelet transform with the discrete cosine and marking full use of the characteristic of the two transforms.Experimental results demonstrate that the algorithms are robust and imperceptive at the attack of noise stain, low-pass filtering and sampling again.