倒谱域自适应音频水印算法  

Robust Adaptive Audio Watermarking Algorithm in Cepstrum

作　　者：年桂君[1] 王树勋[1] 唐笑年[1] 

机构地区：[1]吉林大学通信工程学院

出　　处：《吉林大学学报（信息科学版）》2008年第1期55-61,共7页Journal of Jilin University（Information Science Edition）

摘　　要：为了更好地平衡水印系统的鲁棒性和不可感知性,基于人类听觉感知特性提出了掩蔽度的概念,依据音频信号掩蔽度的大小,选取嵌入帧和调整嵌入强度。针对音频信号倒谱变换后倒谱系数的特征,采用统计均值的调制方法,实现了水印的嵌入位置和嵌入强度两方面的自适应嵌入。在保证水印的不可感知性的情况下,鲁棒性得到了更大的提高。仿真实验证明,该水印算法能经受住包括时间缩放(Time-Scale Modification)和随机剪切(Random Cropping)两种具有挑战性时域同步等大部分攻击,表现出很好的鲁棒性。与SMM (Statistical Mean Manipulation)算法相比,比特误码率BER(Bit Error Rate)至少提高0.25%,随机剪切3 (5×1000)提高了28.5%。To trade-off imperceptibility and robustness, selecting embedding frames and changing embedding intensity adaptively is an effective method. A definition of masking level based human hearing system is given. Embedding frames are selected and intensity is changed based on masking level. According to the feature of cepstrum coefficients, tatistical-mean manipulation method is applied. It is an adaptive embedding algorithm both in embedding location and intensity. Experiment results have shown that the algorithm is robust against most attacks including challenging time-scale modification and random cropping. Compared with SMM （Statistical Mean Manipulation）, BER （Bit Error Rate） was improved at least 0.25%, and the random cropping 3 （5 × 1 000）was improved 28.5%.

关 键 词：音频数字水印 掩蔽度 倒谱变换 统计均值调制方法 

分 类 号：TN912.3[电子电信—通信与信息系统]