Research On Audio Digital Watermarking Based On Discrete Wavelet Transform

Along with the rapid development of multimedia technology and computer networks, the digital media has become a main way of information communication. However, the currently rapid development of digital signal processing and Internet transmission techniques have threatened the copyright of digital data severely. Therefore, how to protect the copyright of digital products in an open network environment effectively becomes an attractive topic. As an important part of information security technology, digital watermark is an effective solution to the protection of copyright. This technology embeds some non-perception information into some carrier data, and the information can be extracted as a proof of copyright. In addition, the digital watermark can be used in secret communication, broadcast monitoring, information identity, and also access control. Therefore, research on digital watermarking is of great significance.At present, research and application of digital watermark mainly focus on image watermark and video watermark. Because of the sensitivity of the human audio system, it is difficult to embed watermark in audio signals, so audio digital watermarking was selected as a research topic in this thesis. By studying the three important features of digital watermark, as well as the features of the human audio system, the discrete wavelet domain is selected to embed watermark among many audio watermarking algorithms.In this thesis, first, a method based on additive embedding in low frequency coefficients of discrete wavelets transform is designed, which can be easily realized, and is of robustness against several attacks. Second, since there are some shortcomings in additive embedding algorithm, such as non-blind extracting, fixed embedding strength and poor security, a method based on self-adaptive bipolar quantization with Logistics sequence encrypting is designed, which is a blind algorithm that watermark can be extracted without original audio signals. Furthermore, taking advantage of human audio system features in frequency domain masking, a method which embeds watermark in high frequency coefficients of discrete wavelets transform is also designed, it is inaudible and robust against several attacks.