【作者】 郑兰君；

【导师】 张秋余；

【作者基本信息】 兰州理工大学 ， 通信与信息系统， 2010， 硕士

【摘要】 随着网络音频文件的广泛传播,以及信息安全领域面临的严峻形势,音频隐写技术已经受到越来越多的关注。本文在分析总结音频隐写技术当前的研究现状与存在的问题后,主要围绕在保证不可感知性和鲁棒性的前提下,提高隐秘信息嵌入容量问题展开研究,在小波域设计了两种不同的音频隐写算法。首先,针对传统直接相加、量化编码,以及修改能量比等嵌入方法存在的嵌入量难以提升的问题,引入矩阵编码的嵌入方法。该方法的特点是能够大幅提高嵌入效率,从而达到提高嵌入量的目的。并且矩阵编码有助于减小对原始音频的修改比例,进而减小嵌入数据对原始音频载体不可感知性的影响。为了改善传统一代小波运算量大,执行速度慢的问题,算法选择提升小波变换系数来嵌入数据。考虑到要保证保密语音的绝对隐蔽,算法利用MPEG I心理声学模型1来计算掩蔽阈值,由掩蔽阈值来控制嵌入帧。其次,针对以往嵌入方法对小波系数浮点数直接处理存在的难以精确控制的问题,引入PCM量化编码将小波系数分段均值量化为二进制序列。该方法可极大减小量化误差,比传统普通单值量化法具有更强的鲁棒性。算法仍然利用矩阵编码的嵌入方法,是对每一个小波分段均值而言,根据应用需求设定分段大小,则可实现嵌入容量的提高。为了降低运算量,算法利用复杂度较低的音频信号时频分析,通过寻找浊音帧来选择嵌入帧,不可感知性亦能得到保证。通过实验分析,本文提出的两种小波域音频隐写算法较同类算法具有较高的嵌入容量,并保证了良好的鲁棒性和不可感知性。更多还原

【Abstract】 With widely disseminating of network audio document and the serious situation in the information security field, the audio steganography have been under more and more attention. This paper analyzes and summarized the current research status and existing problems of the audio steganography, then the major focus is around ensuring imperceptibility and robustness to enhance the capacity of hidden information embedded, and designs two kind of audio steganography algorithm in wavelet domain.Firstly, according to the problem that some embeding methods such as plusing directly, quantization coding, modifing the energy ratio and so on are difficult to enhancing the embeding capacity, so the matrix coding is introducted. This method’s characteristic is capable of greatly improving the embeding efficiency, so as to achieve the purpose of raising the embeding capacity. And the matrix coding help to reduce the proportion of the original audio changes, thereby reducing the effects of the original audio carrier’imperceptibility when embeded datas. In order to improve the problems as large calculation amount and slow execution speed of the traditional generation wavelet, the algorithm chooses the lifting wavelet transform coefficients to embed datas. Considering ensuring absolute concealment of the secret speech, the algorithm uses MPEG I Audio Psychoacoustic model 1 to compute masking threshold, and controls the embedded frames by masking threshold.Secondly, according to the previous embeding methods’s problem of the difficult controling to the wavelet coefficient’s floating point numbers caused by processing directly, the PCM quantization coding is introducted to quantify the every segment’s mean value of wavelet coefficient to a binary sequence. This method can greatly reduce the quantization error, and have stronger robustness than traditional ordinary single-valued quantization. The algorithm still uses the matrix coding embeding method, but it’s for every segment’s mean value of wavelet coefficient. Section size set according to application needs, you can achieve the embedding capacity to improve. In order to induce the calculation amount, the algorithm uses the less complex time-frequency analysis of audio signals to select the embedded frame by looking for voiced frame, imperceptibility can also be guaranteed.Through the experimental analysis, the two kinds of wavelet domain audio steganography algorithm presented in this paper have higher embeding capacity than the same kind of algorithms, and guarantee a good robustness and imperceptibility.更多还原