矢量传感器声源方位与噪声协方差阵联合估计  

Joint Estimation of Sound Source Location and Noise Covariance Using an Acoustic Vector-sensor

作　　者：金勇[1] 程云志[1] 胡振涛[1] 

机构地区：[1]河南大学图像处理与模式识别研究所,河南开封475004

出　　处：《河南大学学报（自然科学版）》2014年第4期467-473,共7页Journal of Henan University:Natural Science

基　　金：国家自然科学基金资助项目(U1204611);河南省基础与前沿研究项目(132300410278);河南省高校青年骨干教师项目(2010GGJS-041)

摘　　要：针对空域非均匀白噪声,通过矩阵变换估计噪声协方差矩阵,进而采用噪声预白化技术获取声源方位的最大似然估计,该方法避免了原最大似然算法加权参数选取的一维搜索过程,提高了算法效率;针对空域色噪声,利用噪声协方差矩阵的先验分布信息,结合噪声预白化技术,提出声源方位与噪声协方差矩阵迭代联合估计算法.仿真实验表明,空域非均匀白噪声背景下,与原最大似然估计方法相比,方位估计精度基本相同,但算法效率更高;空域色噪声背景下,与原最大似然估计方法相比,该方法提高了方位估计精度.This paper studies the maximum likelihood DOA estimation of sound sources based on acoustic vector sensor in two different cases. In the first case, where non-uniform spatial white noise is present, noise covariance matrix is first estimated by matrix transformation and then the maximum likelihood estimation of the sound source is obtained by noise pre-whitening technology, which avoids the one-dimensional search in the parameter selection process of the original maximum likelihood method and improves the efficiency. In the second case, where spatial colored noise is present, the prior distribution information of the noise covariance matrix and noise pre-whitening technology are combined to produce the sound source DOA and noise eovariance matrix iterative ioint estimation algorithm. Simulation results show that, compared with the original maximum likelihood estimation method, in the case of non-uniform spatial white noise, the proposed method has similar estimation accuracy but higher efficiency; in the case of spatial colored noise, the proposed method improves the estimation accuracy.

关 键 词：矢量传感器 噪声协方差阵 最大似然 方位估计 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]