Research On Robust Audio Digital Watermarking Scheme In Transform Domain

The rapid development of Internet and multimedia technology has provided great convenience for the use and dissemination of audio media,but the accompanying information security problem has also become a global problem that needs to be solved.Digital watermarking technology can be used to prove the ownership of audio content or transmit secret information by embedding marker watermark into audio media,so it has been widely used in copyright protection,covert communication,identity authentication,content annotation,military intelligence and other fields.Research on audio watermarking technology has become a hot topic in the field of communication and information security in recent years.Research on robust audio watermarking mainly focuses on enhancing payload capacity and resisting external attacks without affecting the usage value of the host audio,so as to achieve the purpose of transmitting secret information and protecting the ownership of the audio media.In the application of transmitting secret information,the payload capacity,robustness,security and recovery of the extracted confidential information are emphasized.The application of ownership protection focuses on the robustness to combat multiple malicious attacks.At present,most robust watermarking algorithms are only used to hide image or sequence watermarking,but not suitable for audio watermarking with large amount of data and extremely sensitive to BER.Moreover,there are some problems,such as being unable to resist malicious attacks,low hiding capacity,poor transparency,lack of effective synchronization mechanism and no effective measures to enhance the quality of the extracted information.This paper focuses on the robust audio watermarking algorithm used in the above two aspects,including the enhancement of the hiding capacity,robustness,security and noise elimination of audio watermarking.The main research results include the following aspects:(1)In view of the shortcomings of most audio watermarking algorithms,such as low hiding capacity,poor robustness and lack of effective quality enhancement measures,a novel audio watermarking algorithm is proposed based on wavelet transform.The algorithm modifies the high-frequency wavelet coefficients in each audio segment,and then changes the energy values of the two packages in each audio to hide confidential information according to the characteristic that the human auditory system is insensitive to small changes of wavelet coefficients.When extracting information,blind extraction of information can be achieved because that the binary values can bejudged only by comparing the energy values of the two packages of the wavelet coefficients.Before the secret information is embedded into the host carrier,it will be precrypted to prevent information leakage.The embedding depth,hidden frequency band and segment length of the host audio have significant effects on the payload capacity,SNR of the host audio and BER of the extracted information.In practical application,the optimal parameters can be configured according to the actual requirements.Experimental results show that the algorithm has good transparency,high security,large payload capacity and strong robustness for resisting the common attacking behaviors,such as white noise,low-pass filter,echo interference,re-sampling and re-quantitative,and it can be used to hide arbitrary binary data.In particular,the proposed denoising method can effectively remove impulse noise in audio watermarking and enhance its auditory quality.(2)Aiming at the problem that the audio watermarking technology is difficult to resist various attacks because of the poor robustness in practical application,a robust audio watermarking algorithm is proposed in the DWT-DCT hybrid domain.Human auditory system is insensitive to the minor changes in the frequency components of the audio signal,so the watermarks can be embedded by slightly modifying the frequency components.The algorithm uses DWT to decompose the audio signal into wavelet coefficients with different frequency bands,and then makes use of DCT to compress the energy of the specific wavelet coefficients.According to the design embedding rules,the two states of binary watermark are represented by two embedding depths with different values,and then the binary watermark can be embedded into audio media.When extracting watermarks,the embedding depth of the audio fragment is calculated firstly,and then the watermark bit can be extracted by comparing the embedding depth without original audio.Experimental results confirm that the proposed algorithm has good imperceptibility,large payload capacity and strong robustness when resisting against MP3 compression,low-pass filtering,re-sampling,re-quantization,amplitude scaling,echo addition and noise corruption.(3)An adaptive robust audio digital watermarking algorithm is proposed to improve the overall performance of the watermarking algorithm based on DWT and DCT.This algorithm designs new watermark embedding and extraction rules by comparing the average amplitude of the two transform domain coefficients obtained by DWT and DCT.After analyzing the relationship between the algorithm performance and the embedding depth based on watermark embedding and extraction rules,an adaptive control strategy is proposed to control the embedding depth of each audio segment by the overall average amplitude of the audio segment to effectively improve therobustness and imperceptibility.The watermark is encrypted by a chaotic sequence to enhance the security of watermark,so only users who hold the correct key can accurately extract the watermark without the original audio signal.Experimental results show that the proposed algorithm has larger capacity,higher imperceptibility,better security,and stronger robustness when combating against signal-processing attacks than the previous algorithm and involved audio watermarking algorithms in recent years.(4)An audio watermarking algorithm based on DCT and singular value decomposition(SVD)is proposed to overcome synchronous attacks such as time scale modification,pitch shifting modification,random cropping and jittering which will lead to difficult extraction and even loss of the watermark.This algorithm utilizes several measures,such as "secondary uniform frame","local maximum tracking algorithm" and "repeat three times to embed the same watermark" to effectively enhance the robustness against synchronization attacks and conventional signal processing attacks.Based on the analysis of the characteristics of the synchronization attack,a local maximum tracking algorithm based on voiced tone is proposed to search the best watermark embedding position in audio.After DCT transform is performed on the time-domain data in the best embedding region,the medium frequency coefficients of DCT are divided into blocks and SVD is used to decompose them.Finally,the watermark embedding rules are designed by quantization.The security of the algorithm is enhanced by chaotic encryption before the watermark information is embedded.Blind extraction of watermarks can be achieved without the participation of original audio.The experimental results demonstrate that the proposed algorithm has strong robustness against most synchronization attacks and signal processing attacks,large payload capacity and good imperceptibility compared with other algorithms in related literature.