The Audio Watermark Nonlinear Detector Research

With the digital media resources and internet applications, digital information can easily be edited, copied and disseminated without restriction, it is urgent to solve the problem of the piracy and copyright disputes. Since the traditional encryption technology can only be applied to certain problems, and has weak security and poor circulation, digital watermarking as an important approach of copyright protection has attracted wide attentions. The first application of digital watermarking is in images, and then its theories and technologies are improved gradually. Later, digital watermarking has been widely used in audio and video. Digital audio watermarking as an emerging field, there are still many problems to be solved.This paper studies a non-linear detector for detecting the audio watermark. First, the discrete cosine transformation (DCT) coefficients for the audio signal are computed, and reordered by their absolute amplitude. Then the watermark message is embedded into a selected set of DCT coefficients in the median frequency domain. Based on Gaussian mixture probability density function model of the marked DCT coefficients, a non-linear detector is designed for watermark detection. Experimental results show that this kind of non-linear detector has better robustness against additive Gaussian noise, re-sampling, re-quantization, MPEG compression, Chebyshev low-pass filter, and so on. Through the research of nonlinear detector in this paper, it has reference value to find a optimal detector in the non-Gaussian noise environment.