Digital Audio Watermarking Research Base On Intelligent Algorithms

The rapid development of information technology and network technology can not only provide greatly facilitate for the storage and dissemination of multimedia data (audio, image, video, etc.), but also improve the efficiency and accuracy of the information expression. At the same time, the multimedia digital information can easily be conducted even if it is unauthorized dissemination, reproduction and editing for it's convenience. The negative effects, caused by the features of replicability and broad dissemination, of digital technology have become one of the major obstacles which has a great impact on the healthy and sustainable development of information industry. The information hiding technology, represented by digital watermark technology, has become a new certified information security technology, it has effectively realized the copyright protection and content authentication of multi-media information. Although the digital audio watermark technology can not stop the occurrence of piracy. First, it can judge whether the object is protected or not. Second, it can monitor the spread, the judgement of ture or false and the illegal copy of the protected object. Third, it can resolve copyright disputes, and provide evidence for court. It can play an important role in attacking illegal copy and protection of intellectual property rights.Since the digital watermark has been proposed, it has made a great progress and it has also achieved practical using and commercialization to some degree. However, there are also many deficiencies of some representative watermark. On the one hand, some of them lack of robustness. On the other hand, they can not find balance between the robustness and inaudibility.This paper proposed a robust blind audio watermark method base on researching and analysising of many robust audio watermark algorithm and the characteristics of speech signal.This method effectively solves the contradiction of robustness and inaudibility. The main task of this paper can be shown as follows:(1) This paper proposed a new robust watermark algorithm base on SVM in wavelet domain. Watermark signal adaptively embed in low-frequency sub-band coefficients of DWT domain to build a relational model. Under watermark's binary characteristic, the algorithm extracts the watermark into a support vector machine classification model by adding the template information. The use of support vector machine in the case of small samples with good generalization performance of learning and the characteristics of effective realization of the robustness of the watermark signal detection.(2)Through the analysis of audio signals, combined with psychoacoustic model, this algorithm optimize part of location of the embedded with non-perception. The introduction of genetic algorithm, considering the design of robust and non-perception of fitness function, to optimize the location of choice for the overall search to get the final position of the watermark embedding. This algorithm good solution to solve the robust and non-perception of the contradictory requirements of both sides.