基于高阶统计量和功率倒谱的语音回声隐藏分析  

Steganalysis for Speech Echo Hiding Based on High Order Statistics and Power Cepstrum

作　　者：励书怀[1] 汪云路[1] 

机构地区：[1]杭州电子科技大学通信工程学院,浙江杭州310018

出　　处：《大众科技》2012年第7期1-4,6,共5页Popular Science & Technology

摘　　要：文章提出了一种基于功率倒谱和结合两种高阶统计量——偏度和峰度的语音回声隐藏盲检测方法。根据回声隐藏时对载体语音引入回声的特点,通过分析语音的功率倒谱,提取并结合语音功率倒谱的高阶统计量——偏度和峰度。通过对原始语音样本和藏密语音样本功率倒谱偏度-峰度值的学习和训练,提取得到偏度-峰度阈值,可用于回声隐藏的盲检测分析。实验结果表明,该方法能够对四种回声隐藏方法同时生效,且检测率较高。An improved steganalysis method for speech echo hiding based on power cepstrum and high order statistics is proposed.According to echo in stego speech,power cepstrum coefficient of speech is analyzed and its high order statistics including coefficients of skewness and kurtosis are computed in combination.By training the combinatorial skewness and kurtosis in power cepstrum of pure speech and stego speech,we obtain combinatorial skewness-kurtosis thresholds,which can be used for blind steganalysis of echo hiding.Experimental results show that our proposed steganalysis does well in detecting 4 kinds of echo hiding.

关 键 词：回声隐藏 隐藏分析 功率倒谱 偏度 峰度 

分 类 号：TP309.7[自动化与计算机技术—计算机系统结构]