摘要
本文提出了一种基于多重统计量分析的小波域语音信息隐藏算法。该算法首先将载体语音信号分成若干包含相同采样点的帧,利用短时能量以及过零率找出属于浊音段的帧分别进行多尺度离散小波分解,提取小波分解后的低频系数;然后对低频系数进行分组并计算各组系数的能量、绝对值方差等统计量的值,根据各组统计值的比较及嵌入的秘密信息比特值,采用不改变或者适当调节各组统计值大小关系的方法来隐藏信息。该算法只在语音的浊音段嵌入信息,充分考虑了人耳的听觉特性。实验结果表明:算法可以盲检测,对加噪、低通滤波、重采样、重量化等攻击均具有良好的稳健性。
This paper describes a robust information hiding algorithm which embeds secret information in wavelet domain of speech signal. At First, voiced segments is detected from original speech and those frames belong to it are transformed by wavelet decomposition ,withdrawing low frequency coefficients of the transformation. Then the coefficients of each frames are separated into two groups, the energy and variance of absolute coefficients value are computed for each group, according to statistics comparison and value of secret information bits, adopting the way of not changing or slightly adjusting the value of statistics to hide information. This algorithm only hides information in voiced speech,so it fully utilizes human auditory characteristic. Experiment results show that this proposed algorithm can be extracted blindly, and it＇s robust to attack of noise,low-pass filtering, re-sampling and re-quantization.
出处
《信号处理》
CSCD
北大核心
2008年第3期500-503,共4页
Journal of Signal Processing
关键词
信息隐藏
小波分解
统计量
语音信号
information hiding
wavelet decomposition
statistics
speech signal
