论文部分内容阅读
为了满足移动语音通信中对语音内容保持操作的鲁棒性和区分性,并实现高效认证等要求,提出一种基于线性预测最小均方差(LP-MMSE)的高效语音感知哈希认证算法.该算法首先对待认证语音进行预处理、分帧和加窗后的语音信号进行线性预测分析(LPC)得到其最小均方差(MMSE)系数矩阵;然后对分帧后的语音利用谱熵法计算得到每帧的谱熵值参数矩阵;最后结合上述两个矩阵,生成最终的二进制感知哈希序列.仿真结果表明,该算法的感知哈希摘要性优于现有的其它方法并对内容保持操作具有较好的鲁棒性和区分性,认证效率高,能够满足语音通信实时认证的要求.
In order to satisfy the requirement of robustness and distinction of voice content maintaining operation in mobile voice communication and to achieve efficient authentication, an efficient speech-aware hashing algorithm based on linear prediction minimum mean square error (LP-MMSE) is proposed. Firstly, the algorithm performs linear prediction analysis (LPC) on the pre-processed speech of the certified speech and the speech signal of the framed and windowed to obtain the MMSE coefficient matrix; then, Finally, the final binarized perceptual hash sequence is generated by combining the above two matrices.The simulation results show that the perceptual hash summary of the proposed algorithm is superior to other existing methods and has a better performance on content preserving operation Good robustness and distinction, high certification efficiency, to meet the requirements of real-time voice communication certification.