Research On Robust Digital Audio Watermarking

In recent years, With the rapid development of computer network and multimedia information technology, digital audio tampering, large-scale duplication, exchange and transmission becomes very easy, which has led to the serious problem of piracy of audios. In addition, network communications security problems are becoming increasingly prominent. By using the time redundancy of the digital audio and the masking effect of the human auditory system, digital audio watermarking technology can protect the copyright of digital audio products, prove the authenticity of the audio products, track piracy, provide additional information of audio products, and conduct covert communication. Through a suitable hiding algorithm, the watermark is embedded into the digital audio, which can not significantly undermine the quality of audio signals. Digital audio watermarking is an important branch of information hiding technology, which is an effective method for copyright protection.The watermark is a weak signal hidden in the audio signal.The strength of the watermark signal should be much lower than that of the audio signal as an information carrier. The watermark extraction can be done by using detection equipment or watermark extraction equipment. In most cases, it is required that the audio watermark be auditorily imperceptible, which means that the embedding of the watermark should not affect the auditory effect of the audio carrier. Imperceptibility (or transparency) is based upon the characteristics of our auditory senses. Compared with embedding watermarks into still images, audio watermarking is significantly more challenging due to the extreme sensitivity of the human auditory system (HAS) to changes in the audio signal.Three important and basic issues that should be considered in watermarking system are:1) perceptual transparency,2) watermark bit rate, and3) robustness against attacks. There are trade-offs between these three defining properties. Normally, the robustness is improved along with the decrease in transparency and capacity; improving the embedding capacity will bring about decreased transparency; to enhance transparency is often to sacrifice robust performance or embedding capacity. Since1993, the concept of digital watermarking has been proposed, audio watermarking research has drawn the attention of scholars, and research results have been published, but there still exist many deficiencies in robust audio watermarking performance, for example, most algorithm does not consider the anti DA/AD transformation. In addition, with the wide application of the streaming media, streaming real-time audio watermarking with traditional audio watermarking has been applied in different=environments.In order to achieve the audio copyright protection, and meet the needs of audio covert communication, first, the present study conducted a review of robust audio watermarking technology, and summarized and synthesized three aspects including conventional robust audio watermarking technology, anti-DA/AD conversion robust audio watermarking algorithm, and high-capacity audio watermarking technology. The most robust audio watermarking technology shows the problem of lack of robust performance. They often can only resist common signal processing such as noise, and low-pass filtering. High-performance robustness of the algorithm is usually very large in computation. Existent robust audio watermarking algorithm resistant to DA/AD conversion is relatively rare. There exist deficiencies in anti-DA/AD algorithm performance. Compared with the ordinary robust audio watermarking research, large-capacity audio watermarking (steganography) algorithm has received relatively less attention. Existing algorithm is often at the expense of the robustness or transparency of the algorithm to enhance the capacity. Research concerning how to improve the robustness or capacity of audio watermarking technology for different applications is very meaningful work.This paper focused on robust audio watermarking technology research, and studied the common views of watermarking attacks, including a variety of signal processing attacks, and the mainstream tool attack and MPEG compression attack. The study put focus on the analysis attack algorithm for changes in audio signal time frequency characteristics. It analyzed the classic watermarking algorithm robust performance in the time domain and transformation domain, and the different characteristics of embedded domain watermarking algorithm performance. For different application requirements, it selected the appropriate embedded domain and balanced perception of transparency, robustness and embedding capacity of the three requirements, considered the characteristics of a variety of attacks, and designed embedded algorithms and extraction algorithm. The present research presents a strong robust performance in the conventional application of audio watermarking algorithm, anti-DA/AD conversion robust audio watermarking algorithm and large capacity robust audio watermarking (steganography) algorithm. Therefore, this study is of great significance for copyright protection of digital audios and covert communication applications.In this paper, the major contributions are as follows:Firstly, a robust audio watermarking technology based PPG transformation was presented.In conventional audio copyright protection, the traditional robust audio watermarking algorithm design is directly based on one-dimensional linear audio signal. Algorithm [89] creatively put forward the time-domain quantization blind watermarking scheme based on two-dimensional space, but the program assumed that the audio sampled Gaussian distribution, which might not be able to meet the actual situation. In view of this issue, drawing on the field of multimedia high-dimensional thinking, and by analyzing the statistical characteristics of wavelet coefficients and the watermarking algorithm performance, the strong robust audio watermarking algorithm was proposed.The algorithm selected wavelet domain as an embedded domain.Firstly, the discrete wavelet transform sub-band coefficients are appropriate morphological transformation (PPG transform).Then, based on the relationship of the sizes of the adjacent two statistical moments embedded algorithms and extraction algorithm were designed. Finally, according to simulation optimization the parameters of the algorithm were adjusted. Three-dimensional morphological transformation takes advantage of the short-time correlation of the audio signal, the noise introduced by the watermark is distributed to the broader region,where is not sensitive to the human ear.In the same circumstances, the algorithm is better than traditional algorithm on transparency, performance robustness, and the correct watermark extraction rate.Simulation results showed that the algorithm could effectively resist common signal processing attacks, such as filtering, noise, weight, MPEG compression and the mainstream assessment tool Stirmark for Audio. The algorithm watermark extraction rate was over90%. The objective auditory quality of the proposed algorithm was more than30dB.Secondly, based on Chirp signal and the Radon transform, Anti-DA/AD conversion robust audio watermarking algorithm was designed.In order to design a robust audio watermarking that can be used in streaming media, concerts and television broadcasting and other applications, some research has been conducted. After the analysis of the DA/AD conversion process, the damage was obtained that it brings to the audio watermarking algorithm. We have come to the conclusion that the DA/AD conversion will result in the audio signal waveform distortion and phase distortion.Based on Chirp signal and the Radon transform characteristics analysis, an anti-DA/AD that transforms robust audio watermarking algorithm was proposed.The algorithm selected wavelet domain as an embedded domain to calculate the adjacent frame signal statistical moments of the Radon transform, with a moment magnitude relation to characterize the embedded bit. Meanwhile chirp signal with a specific frequency characteristics was introduced in order to improve the robustness of watermarking algorithm anti-DA/AD conversion performance. Because the Chirp signal has strong penetration ability, good self-correlation and noise immunity characteristics as well as the Radon transform geometric invariant for translation, scaling and rotation, so the algorithm can have a good anti-DA/AD conversion performance.In the DA/AD conversion cable-based transmission simulation environment, and the conditions of the different players and recording softwares (such as Windows Media Player, STORM, Cool Eidt, etc.), the experimental results showed that the proposed watermarking algorithm correct extraction rate was more than90%.Thirdly, large-capacity audio steganography algorithm was based on source separation. This part of the study was aimed to meet the needs of covert communication and high-capacity watermarking applications. In order to improve the steganography or watermarking capacity, a new type of high-capacity audio steganography was proposed by studying the existing source separation technology, and based on a sound source separation technology. The covert audio signal was separated into estimated source signals by using source separation algorithm and suitable steganography method. Then the hidden source audio was analyzed into stego audio through special mixture method. The test results show that, compared with the classic hidden technology,If we choose N isolated the source, using the classical algorithm for watermark embedding, we can get the watermark capacity is N times that of classical algorithm.In summary, the present research studied the robust performance and embedding capacity of robust audio watermarking technology in different application environments. It proposed three robust audio watermarking technology,which could be applied to different environments. The study has important theoretical significance and practical value. On this basis, the paper concludes research innovations, and looks forward to further research combined with the existing research results against security attacks.