Efficient Audio Zero-watermarking Algorithm For Copyright Protection Based On BIC And DWCM Matrix

With the rapid development of Internet and multimedia processing technology, ithas become increasingly important to protect the copyright of the digital worksincluding multimedia. A widely studied algorithm for digital audio copyright protectionis based on digital watermarking technology. In this paper, to make the watermarktechnology meet the requirements of the audio copyright protection, a novel audio zero-watermarking scheme based on BIC and DWCM matrixes is proposed. The audiowatermarking algorithm can avoid the contradiction between strong robustness andtransparency, and resist the attack of large-area random cropping, and solve thewatermark synchronization problems.In the thesis, the main research contents and innovations are as follows:①This paper presents a new not-embedded digital audio watermarking scheme,the digital zero-watermarking algorithm. The digital zero-watermark algorithm isdifferent from the conventional embedded watermark algorithm. In the watermarkembedding process, watermark information is not really embedded to the host audio,but mixed into the audio characteristic matrix to get a key by XOR. In the watermarkdetection process, the key is used as the password of the watermark image extraction ordetection. The idea of digital zero-watermark solves the contradiction between thestrong robustness and transparency perfectly.②An audio segmentation algorithm based on Bayesian Information Criterion(BIC) was introduced to the study. Against the two shortcomings of the conventionalaudio algorithm based on BIC, the algorithm was improved to eliminate theaccumulated error and solve the hard decision. Before the watermarking "embedding"and detection, the using of audio BIC segmentation algorithm solves the defect that theexisting watermarking algorithm cannot overcome large-area random cropping andsolves the watermark synchronization problems③Propose the Double-direction Wavelet Coefficients Mapping (DWCM)Matrixes and introduce how to extract DWCM matrixes. In the process of extractingDWCM matrix, we use a new approach to map the wavelet approximate coefficients ofaudio clips into a binary image matrix. The process of extracting DWCM matrix is thekey step of the algorithm proposed in the study.In the paper, first, the improved BIC segmentation algorithm is used to split theaudio into audio clips with different length and filter fragments including tiny samples. Then, implement wavelet decomposition to each audio clip and extract the DWCMmatrix. Finally, the watermark information is "embedded" or detected from the each clipof audio signal. In the watermark "embedded" stage, implement XOR operation onDWCM matrixes of all audio clips and the watermark image to generate a key set,which is submitted to third-party trusted institutions and regarded as the password of thewatermark detection. In the watermark detection stage, with the keys obtained from thethird party trusted institutions, the watermark image of unknown audio can be detectedby BER matching.The experimental results demonstrate that the proposed zero-watermark schemehas more excellent performance in terms of security, robustness, anti-croppingcompared to the earlier audio watermarking schemes. This is mainly contributed by theproposed BIC audio segmentation and DWCM Matrix. To our best knowledge,zero-watermarking copyright protection based on BIC audio segmentation and DWCMMatrixes has not been reported before.