Research On Digital Audio Watermarking Algorithm Based On Transform Domain

The technology of embedding the owner's effective copyright information in the audio signal is called digital audio watermarking technique.This technology can effectively solve the copyright problem of digital audio,and has become one of the important researches in the field of information security.Synchronization is very important for the research of audio signal,so the development of audio watermarking technique is slow;With the continuous development of Internet technology,a large number of various cracking tools lead to more and more serious infringement of digital audio,Nowadays,the performance of audio watermarking algorithm has higher requirements.However,due to the diversity of music audio types,for the existing frequency domain audio watermarking algorithm,using different transform cascading method to embed watermark information can't guarantee that it is suitable for most types of music audio at the same time,and has no good generalization ability.Therefore,this paper studies and improves the traditional frequency domain audio watermarking algorithm.This paper mainly studies and summarizes from two aspects,one is for the traditional embedded audio watermarking algorithm,the other is for the "non embedded",that is,the traditional audio zero-watermarking algorithm.The work details are as follows:Firstly,this paper introduces the content of audio signal(including speech signal)by consulting the literature and sorting out the data,carefully studies the classic papers and frontier papers in the field of audio watermarking,and systematically summarizes the relevant knowledge of audio watermarking technology.This paper introduces the principles of some classic audio watermarking algorithms,including the audio watermarking algorithm based on frequency domain,analyzes the advantages and disadvantages of different methods,and puts forward the corresponding optimization ideas for the existing problems.Secondly,in the traditional embedded audio watermarking algorithms,the tradeoff impecerptibility,payload and robustness isn't solved effectively.Meanwhile,the watermark information shows weak security.Aiming at these two problems,an audio watermarking algorithm based on secret sharing and stationary wavelet transform(SWT)is presented.Shamir's secret sharing scheme is used to process the watermark information to obtain n pieces of secret information,of which n-1 copies are securely stored in the blockchain,and the remaining one is embedded in the selected voiced frames according to the hash code generated by the watermark information feature.In the frequency domain,by modifying the discrete cosine transform(DCT)coefficients of selected voiced frames through SWT and Schur decomposition(SD),the watermark is adaptively embedded in the first column element of the orthogonal matrix obtained from SD.The experimental results show that the method has high imperceptibility and is strong robust to various signal processing attacks,such as filtering,resampling,re-quantization,MP3 compression and random cropping,the payload reaches1.39 kbps.Thirdly,in order to make the selected audio features more representative in the construction of zero watermark,and remove the accompaniment to participate in the feature extraction process,a robust audio zero-watermarking algorithm based on Support Vector Machine(SVM)and harmonic features is presented.Firstly,the speech signal in the original audio is modeled by the Harmonic Noise Model(HNM)to obtain the harmonic part,then the K-means clustering algorithm is used to obtain the large amplitude region of the high pass SWT sub-band of the harmonic part of each frame to construct the zero watermark.The mean value set of singular values determined by the large amplitude region is used as the SVM sample set for SVM training to generate the decision function.The watermark detection method is blind detection,which uses the decision function obtained in the stage of constructing zero watermark to detect.Experimental results show that the algorithm has a strong anti-attack effect for different types of audio.