论文部分内容阅读
提出一种基于双树复小波变换的多格式音频感知哈希算法,解决了现有音频认证算法音频格式单一、算法不通用、效率低的问题.首先对预处理后的音频信号进行全局双树复小波变换,获得信号的实小波和复小波系数,对它们分别分帧,帧数相同;对实小波系数计算每帧信号Teager能量算子的模值,作为实小波系数的帧间特征,接着对每帧信号再分帧,提取再分帧帧信号的短时能量作为实小波系数的帧内特征;对复小波系数求取每帧信号的熵值作为复小波系数的帧间特征;最后对上述特征分别进行哈希构造,生成感知哈希序列.实验结果表明,该算法对5种不同格式的音频都具有强鲁棒性,且区分性好,效率高,并能实现小范围篡改检测.
This paper proposes a multi-format audio perceptual hashing algorithm based on double-tree complex wavelet transform, which solves the problem that the existing audio authentication algorithm has a single audio format, the algorithm is not universal and inefficient. Firstly, the global double-tree Real wavelet and complex wavelet coefficients of the signal are obtained, which are respectively divided into frames and have the same number of frames. The real value of the Teager energy operator of each frame is calculated as the inter-frame feature of real wavelet coefficients, And then the short-time energy of the sub-frame signal is extracted as the intra-frame feature of the real wavelet coefficient; the entropy value of each frame signal is obtained as the inter-frame feature of the complex wavelet coefficient for the complex wavelet coefficient; finally, the inter- The above features are respectively constructed by hashing to generate a perceptual hash sequence.The experimental results show that the proposed algorithm is robust to all five different formats of audio and has good discriminability and high efficiency and can detect small tampering.