论文部分内容阅读
提出了一种基于梅尔频率倒谱系数相关性的语音感知哈希内容认证算法.该算法提取分段语音的声纹梅尔频率倒谱系数作为感知特征.为提高算法的安全性,算法利用伪随机序列作为密钥,计算得到梅尔频率倒谱系数与伪随机之间的相关度,最后量化相关值并加密生成感知哈希序列.语音认证过程中,采用相似性度量函数来衡量哈希序列之间的距离,同时与汉明距离方法进行了比较.仿真结果表明,该算法对语音内容保持操作,如重采样、MP3压缩等具有较好的鲁棒性,相似性度量函数也对语音篡改检测定位具有较高的灵敏性.
This paper proposes a speech-aware hash content authentication algorithm based on the correlation of Mel frequency cepstrum coefficients, which extracts vocalized Mel-frequency cepstral coefficients of speech as perceptual features.In order to improve the security of the algorithm, the algorithm uses Pseudo-random sequence as the key to calculate the correlation between the Mel-Frequency Cepstral Coefficients and the pseudo-random, and finally quantify the correlation value and encrypt to generate the perceived hash sequence.In the process of voice authentication, the similarity measure function is used to measure the hash The distance between the sequences and the Hamming distance method are also compared.The simulation results show that the algorithm has good robustness to voice content preservation operations such as resampling and MP3 compression, Tamper detection and positioning with high sensitivity.