Research And Application Of Digital Audio Watermarking Algorithm

The development of digital multimedia technology and Internet technologies makes the digital media transmission become very convenient. However, the dissemination about piracy music and the modified of copyrighted content lead the interests of the owners and issuers to be damaged. Therefore, the copyright protection has become the issues to be settled urgently. As a potential solution, digital audio watermarking technology has been widely concerned and becomes a hot academic research.Embedded algorithm and the method to select location of embedded are two key issues in the research of digital audio watermarking technology. How to select audio segments which affect audio quality at least becomes a hot issue. The method to modify original audio to embedded watermark affects audio quality and robustness, so embedded algorithm becomes a focus.According to the above two key points, our study focuses on embedded algorithm and the method how to select embedded location. The main works of this thesis are as follow: (1) An algorithm based on DWT is proposed that embedded one watermark signal in two low-frequency wavelet coefficient. Experiments show that the method have robustness resistance to the attacks like smoothing and compression; (2) For how to choose a suitable location to embedded, an adaptive mechanisms based on SVM was proposed. According to properties of digital audio, select a segments of audio have stable energy, zero-crossing rate as a feature vector and get SVR (Support Vector Regression) training model, training model using SVR to achieve position adaptive selection, combined with quantitative strategies to embedding and extraction. Experiments show that the method can select the audio segments which have high energy and low zero-crossing. Also this method affects the audio quality at least and has robustness resistance to the attacks like MP3 compression smoothing and re-sampling.