基于时频域特征的直接扩频音频隐写分析  

Steganalysis of DSSS Audio Based on Time-frequency Domain Characteristics

作　　者：徐雷[1] 郭立[1] 王翠平[1] 王昱洁[1] 杨帆[1] 

机构地区：[1]中国科学技术大学电子科学与技术系,安徽合肥230027

出　　处：《通信技术》2010年第1期78-80,83,共4页Communications Technology

基　　金：国家自然科学基金资助项目(批准号:60772032)

摘　　要：借鉴语音压缩编码的理论模型与方法,对wav音频预处理之后,在频域分析短时离散傅里叶变换(STFT)域的频谱响应曲线;在时域构造全极点模型,分析预测残差,得出了一组对音频直接扩频(DSSS)隐写高度敏感的时频域联合统计特征。并利用层次向量机(H-SVMs)设计二级分类器,提出了一种新的隐写分析方法,不仅可以检测音频是否进行了隐写,还能进一步检测隐写是在FFT、MCLT域,DCT、MDCT域还是DWT域进行的。实验结果表明,此方法对于多种变换域的数据隐藏方法,平均检测准确率达90%以上。From voice compression coding theory models and based on the pretreatment of way audio, the paper analyzes the spectral response curve of short-time Discrete Fourier Transform （STFT） in frequency domain; constructs all-pole model and analyzes the prediction residual in time domain. Thus a group of time-frequency domain characteristics highly sensitive to Direct Sequence Spread Spectrum （DSSS） audio steganalysis is achieved, and a 2-level SVM is designed according to the theory of Hierarchical SVMs （H-SVMs）. Finally, it proposes a new method of steganalysis, which could not only detect whether the audio is in steganography, but further tests steganalysis in FFT, MCLT domain, DCT, MDCT domain or DWT domain. The experimental results show that the proposed method, for data hiding methods of various transform domains, could achieve an average detection accuracy rate of over 90%

关 键 词：隐写分析 DSSS STFT 预测残差 H-SVMs 

分 类 号：TP391.42[自动化与计算机技术—计算机应用技术]