基于短时自相关的语音感知哈希认证算法

Speech Perceptual Hashing Algorithm Based on Short-term Auto-correlation for Speech Authentication

针对移动语音通信对语音区分性和内容保持操作鲁棒性,以及实时认证的要求,提出了一种基于短时自相关的语音感知哈希认证算法。该算法对语音进行预处理、分帧和加窗,进行短时自相关分析,并结合非负矩阵分解得到特征参数矩阵,利用特征参数矩阵生成感知哈希序列进行语音内容认证。实验结果表明,该算法具有较好的区分性和鲁棒性,能够满足语音通信实时认证的要求。

In order to meet the requirements of discrimination,robustness and real-time authentication for mobile speech communication,a speech perceptual hashing algorithm based on short-term auto-correlation for speech authentication is proposed.Firstly,the algorithm preprocesses the speech by frame and adds window function.Then,the short-term auto-correlation is carried out,and the characteristic parameter matrix is obtained by decomposition with non-negative matrix.Finally,the characteristic parameter matrix is used to extract perceptual hashing sequence for speech authentication.The experimental results show that the proposed algorithm has a better discrimination,robustness and can meet the requirements of real-time authentication in speech communication.