Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > General issues > Security and confidentiality > Encryption and decryption

Blind Audio Watermarking Algorithm Based on Machine Learning

Author XuXiaoJuan
Tutor HeChengYuan;PengHong
School West China University
Course Computer Software and Theory
Keywords Information Hiding Digital audio watermarking Wavelet Transform Copyright Protection Support vector regression Genetic Algorithms Blind detection
CLC TP309.7
Type Master's thesis
Year 2008
Downloads 92
Quotes 2
Download Dissertation

With the rapid development of computer network and multimedia technology, multimedia data is becoming an important source for people to obtain information, and become an important part of people's lives. Thus, an effective tool for the declaration and protection of intellectual property rights as a digital media products, digital watermarking technology since the first in 1993, has caused great concern, at the same time, how to protect multimedia information security has become a hot research topic in the international . In particular, the audio signal of the digital watermarking technology has become one of the hot research in recent years. Cryptography technology to take information security technology can only control the information dissemination process for decoding media data is difficult to control, and therefore can not prevent illegal copying and dissemination of pirates. Begun to attract widespread attention as an effective way to solve the above problem, the digital watermarking technology. To prove it by embedding secret information in the original data the ownership or completeness of the data, in order to suppress piracy of digital works or tampering. Secret information is embedded in digital audio watermarking technique is to vector data (audio signal) in order to achieve the purpose of Disclaimer and protection. Digital watermarking technology is the key to the work domain and embedding strategy selection, to some extent, the work domain and embedding strategy to select the good and bad fundamentally determine the pros and cons of the digital watermarking system. Developed in recent years, the wavelet transform is a new kind of time-frequency analysis method, has a lot of good nature, particularly suitable for the processing of the audio signal. This article is the wavelet transform is applied to the audio signal digital watermarking system, proposed a new digital machine learning method based on a wavelet domain blind audio watermarking algorithm so that it can be between watermark robustness and imperceptibility to find a reasonable balance point, which has good transparency and robustness, practicality and copyright protection of audio works. The main contribution of this paper is as follows: (1) propose a robust digital audio watermarking based on support vector regression (Support Vector Regression, SVR) algorithm. The basic idea of ??the algorithm is the first sub-sampling process on the entire audio signal, and then all the sub audio wavelet transformation is performed, respectively, the watermark signal is embedded into a sub-audio signal threat ū Hay slow significantly less Vitex to push Di stool ? Watermark extracted without the original audio signal. Due to the high correlation between the different sub-audio signal, the wavelet low frequency coefficients corresponding DWT decomposition of the distribution has a similar nature. SVR extracted watermark to take advantage of this high degree of correlation, non-linear approximation capability of embedding and extraction process applications SVR, the establishment of the corresponding template relations between the sub audio signal to be embedded with other sub-audio signal, after training to achieve blind detection. (2) for digital audio watermark robustness and imperceptibility between mutual restraint problem, propose a genetic algorithm (Genetic Algorithms, GA) to solve the the optimal embedding energy optimization watermarking scheme. The genetic algorithm is a random search group evolution to the optimization problem of the objective function for the aforementioned blind wavelet domain digital audio watermarking algorithm based on support vector regression to further explore genetic operators to search in the collection of the embedding strength against attack capability fitness than high individual, resulting in the best embedding strength of an optimization program, in order to achieve the adaptive strategy, the algorithm to find the proper balance between watermark robustness and imperceptibility, which has more good transparency and robustness. Simulation results show that the two methods have robustness and imperceptibility watermarking capacity self-adaptive, does not require the participation of the original audio signal to extract the watermark, and including MP3 lossy compression, low-pass filtering, resampling / weight of attack test is robust, viable digital audio watermarking algorithm. Depending on demand, the copyright protection for digital audio works.

Related Dissertations
More Dissertations