摘要
对以固定嵌入强度和未知嵌入强度的小波域乘性音频水印的优化检测算法进行了研究。利用广义高斯分布对音频低频小波系数建立统计分布模型,并使用统计决策理论知识分别对固定嵌入强度和未知嵌入强度的检测统计量进行确定,利用Neyman-Pearson准则对两种嵌入强度的算法的检测阈值进行计算,并对检测算法进行了仿真。实验结果表明,对于固定嵌入强度与未知嵌入强度的乘水印的盲检测算法,该检测算法的性能比音频水印算法中常使用的相关检测的性能优越。
This paper studies the optimum detection model of muhiplicative audio watermarking with fixed embedded strength and unknown embedded strength in discrete wavelet transform (DWT) domain. The generalized Gaussian distribution (GGD) is chosen to statistically model the wavelet coefficients of low frequency sub-bands coefficients. The detection statistics of two different embedded strengths are derived according to statistical decision theory and the detection thresholds are computed using the Neyman-Pearson criteria. Experimental results show that the performance of the new method is better than that of the related detection methods.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2009年第1期147-151,共5页
Chinese Journal of Scientific Instrument
基金
内蒙古自治区自然科学基金:信息隐藏技术的研究(200711020819)资助项目
关键词
音频水印
离散小波变换
广义高斯分布
拉普拉斯分布
虚警率
audio watermarking
discrete wavelet transform (DWT)
generalized Gaussian distribution
Laplacian distribution
false alarm rate