基于对象传播神经网络的抗TSM攻击音频水印算法  

Audio watermarking algorithm robust to TSM based on counter propagation neural network

作　　者：金文标[1] 戴红亮[2] 

机构地区：[1]杭州电子科技大学理学院,杭州310018 [2]重庆邮电大学计算机科学与技术学院,重庆400065

出　　处：《计算机应用研究》2009年第12期4758-4760,4796,共4页Application Research of Computers

基　　金：重庆市自然科学基金资助项目(2006BB2374)

摘　　要：提出了一种基于对象传播神经网络的抗TSM攻击音频水印算法。利用CPN自学习和自适应的特征,通过自适应改变段长的分段算法,选用具有较强稳定性的小波低频系数方差作为输入向量训练CPN,建立音频特征与水印信号的对应关系,以达到嵌入水印的目的。实验结果表明,该算法对常规音频信号处理和TSM等同步攻击具有很强的鲁棒性。This paper proposed a novel CPN-based audio watermarking algorithm against time-scale modification. Utilizing the self-learning and self-adaptive capabilities of CPN and with adaptively changing the length of segment, the relationship between the important characters of audio signals and embedded watermark was learned by using the variance of low frequency wavelet coefficients with strong stability as the input of CPN, to achieve the purpose of embedding watermark. Experimental results show that the algorithm is very robust to common audio signal processing and synchronization attacks, such as TSM.

关 键 词：数字音频水印 对象传播神经网络 时间缩放 小波低频系数方差 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]