面向立体声的多源音频信息隐写

面向立体声的多源音频信息隐写 A Multi-Source Steganography for Stereo Audio 常迪 CHANG Di1 1 2 first-author 张霞 ZHANG Xia3 3 吴越 WU Yue1 1 2 武汉大学国家多媒体软件工程技术研究中心 武汉大学国家多媒体软件工程技术研究中心 National Engineering Research Center for Multimedia Software,Wuhan University National Engineering Research Center for Multimedia Software,Wuhan University 武汉大学计算机学院 武汉大学计算机学院 School of Computer,Wuhan University School of Computer,Wuhan University 江汉大学物理与信息工程学院 江汉大学物理与信息工程学院 School of Physics and Information Engineering,Jianghan University School of Physics and Information Engineering,Jianghan University 针对多音源的音频信号,提出面向立体声的多源音频信息隐写算法.该算法将音源分离技术中的二进制时频掩码技术引入到信息隐写算法中,以各分离音源为载体,采用立体声混合嵌入的方式,通过修改时频域中奇数窗的部分系数,实现隐写信息的嵌入,该算法代替了传统信息隐写算法的单源嵌入方式.实验结果显示,该算法嵌入容量较大,并能有效抵抗常见信号处理攻击、StriMark Benchmark for Audio等攻击,平均误码率小于2.80%,算法具有很好的鲁棒性. We proposed a steganography algorithm for stereo multi-source audio by studying sound source separation technique based on spatial clues and using it in information steganography technology.In this algorithm,the binary spectrum mask technology which is used in sound source separation technology is introduced in steganography algorithm.Regard the separated sound sources as carriers,and embed the stereo sound,by modifying part of the coefficients in odd windows in time-frequency domain,to achieve steganography information embedding,which replaces the traditional single source embedding steganography algorithms.Experimental results show that this algorithm has large embedding capacity,and it is able to resist the attacks of StriMark Benchmark for Audio and other common attacks effectively.The average bit error rate is smaller than 2.80%,which provides good robustness. 立体声音频 信息隐写 时频域 多音源 stereo audio information steganography time-frequency domain multi-sources 10.14188/j.1671-8836.2013.03.011 TN912.3 国家大学生创新创业训练项目（111048652,1210486060）;湖北省教育厅科学技术研究重点项目（D20123405）;湖北省建设科技研究项目（鄂建文2012-78）;武汉市教育局教研项目（2011055） 2013-03-01 2021-04-01 3 59