一种基于谱相减法的声信号时延估计  

Time Delay Estimation for Acoustic Source Based on Spectral Subtraction

作　　者：张慧敏[1,2] 柴毅[1] 周鑫[2] 

机构地区：[1]重庆大学自动化学院,重庆4000044 [2]重庆电子工程职业学院通信系,重庆401331

出　　处：《世界科技研究与发展》2011年第4期641-642,674,共3页World Sci-Tech R&D

基　　金：国家自然科学基金(60974090);教育部博士点基金(200806110016)

摘　　要：利用信号增强算法,结合广义互相关时延估计原理,提出一种声信号时延估计算法。在噪声背景下先采用谱相减法从带噪信号中估计出原始声信号,提高信噪比,利用四元立体传声器阵列和广义互相关算法原理,对相关数据进行加权处理得到较高精度的时延估计值。该算法可以有效抑制空间加性噪声,得到有效声信号准确的时延估计。实验仿真结果表明了该算法的可行性,在信噪比较低的情况下,与单一使用广义互相关函数相比,极大地提高了时延估计算法的准确程度。Adopting signal enhancement algorithm, combined with generalized cross-correlation delay estimation theory, a delay estimation algorithm for acoustic source is proposed. In the noise background, speetral subtraetion is used to estimate original acoustic signal from the signals with noise first, whieh improves signal to noise ratio. Then four elements three-dimensional microphone array and generalized eross-eorrelation algorithm are used to process data to get time delay estimates of high accuracy. With this algorithm, spaee additive noise can be effectively suppressed, with the accurate delay estimation of valid acoustic signal gotten. Numerical analysis and experimental results show the feasibility of the algorithm. In the ease of low SNR, compared to the algorithm which only uses generalized cross -correlation function, the accuracy of time delay estimation algorithm is improved greatly.

关 键 词：时延估计 广义互相关 谱相减法 传声器阵列 

分 类 号：TN912.3[电子电信—通信与信息系统]