摘要
提出了一种用于提取音频水印的符号检测器并研究了其鲁棒性能。首先,对于宿主音频的离散余弦变换系数进行绝对值排序,将水印所加的中频系数段分布假设为对称高斯混合分布模型,并依据这种模型设计了一种局部最优检测器,即符号水印检测器。其次,对于这种符号检测器的检测性能进行了详细理论分析和实验证明。数值实验表明,这种检测器对于高斯噪声、重采样、重量化、MPEG压缩、低通滤波等攻击表现出较好的抗攻击性能,同时与符号检测器性能的理论分析相符。
A sign detector for extracting the audio watermark is proposed and its robustness is studied in this paper.Firstly,the discrete cosine transform（DCT） coefficients of audio signal are sorted by their absolute values.Then,we assume the median frequency domain of the DCT coefficients to be embedded by the watermark as a symmetrical Gaussian mixture distribution model.Based on this distribution model,a locally optimum detector,i.e.,a sign detector,is deduced,and its detection characteristics are analyzed theoretically and experimentally.Experimental results show that the sign detector has a good robustness against such attacks as additive Gaussian noise,resampling,requantization,MPEG compression,and Chebyshev low-pass filter,etc.The experimental results accord to the theoretical analyses of the performance of the sign detector well.
出处
《青岛大学学报（工程技术版）》
CAS
2011年第2期32-37,共6页
Journal of Qingdao University(Engineering & Technology Edition)
基金
山东省自然科学基金资助项目(ZR2010FM006)
关键词
音频水印
对称双高斯混合模型
符号检测器
攻击
audio watermark
symmetrical Gaussian mixture distribution
sign detector
attacks
