基于统计特征的语音回声隐藏分析

Steganalysis of Speech Echo Hiding Based on Statistical Characteristics

提出了一种基于偏度和峰度的语音回声隐藏分析方法。通过分析回声隐藏导致的语音复倒谱变化,计算出语音复倒谱的统计特征——偏度和峰度。通过对原始语音样本和藏密语音样本复倒谱偏度和峰度的学习和训练,分别得到可用于回声隐藏分析的偏度和峰度阈值。该方法已在计算机上进行了实验。实验结果表明,该方法能够有效地分析多种回声隐藏方法,峰度的隐藏分析结果好于偏度。

A steganalysis method for speech echo hiding based on the statistical characteristics of skewness and kurtosis is presented. After calculating these two parameters, the change of the complex cepstrum caused by the echo hiding is analyzed. By learning and training the statistical characteristics of both the pure and stego speeches, the thresholds of the skewness and kurtosis are obtained. The method is tested on a computer. Experimental results show that the proposed mehtod can effectively analyze a variety of echo hiding methods and the kurtosis steganalysis results are better than that of the skewness.