Digital Watermark Detection Method Based on Multi-parameter Weibull Statistical Modeling

Abstract

The invention discloses a digital audio watermark detection method based on multi-parameter Weibull statistical modeling. The watermark embedding is to first select the SWT high-frequency sub-band with larger host audio energy as the optimal sub-band for watermark embedding. Its coefficients are modified, and the subband of the modified coefficients and other subbands are combined to perform SWT reconstruction to obtain watermarked audio; watermark extraction is to perform SWT on the watermarked audio at the receiving end, and use multi-parameter Weibull distribution to its high-frequency subbands. Carry out modeling with important coefficients, and use the correlation of coefficients in subbands to estimate the parameters of the model; finally, use the maximum likelihood and local maximum potential energy test methods to design a local optimal detector to extract specific watermark information, and obtain the final watermark information in order. watermark sequence.

Description

technical field [0001] The invention belongs to the technical field of copyright protection of digital audio, relates to a digital audio watermarking method based on a statistical model, in particular to a digital watermark detection method based on multi-parameter Weibull statistical modeling. Background technique [0002] In today's information age, people can more easily obtain digital multimedia resources such as audio, video and images through mobile devices and Internet technology. As an important part of multimedia resources, digital audio is becoming more and more easy to be illegally copied, tampered with and disseminated, which makes the information security of audio more serious. In order to protect the legitimate rights and interests of audio copyright owners, it is especially important to add identity verification information in audio, and digital audio watermarking technology is one of the most effective methods to achieve this purpose. [0003] In recent year...

AI Technical Summary

Problems solved by technology

However, watermarking methods based on statistical models still have the following shortcomings: first, the transform domains currently used are usually downsampled wavelet transform, discrete cosine transform, and Fourier transform, which cannot capture important information and features of audio well; second , when modeling the transform domain coefficients, there is no in-depth analysis of the selected distribution model, and it is not fully proved whether the established model is optimal for the selected transform domain; third, when the model distribution parameters are estimated, the watermarked audio signal is directly Parameter estimation is greatly affected by the watermark embedding strength; fourth, most of the existing methods use the log likelihood ratio to construct a maximum likelihood detector, and do not try to use other statistical testing strategies to construct a new detector to improve watermark detection. precision

Images