基于对角减载的水声阵列SMI-MVDR空间谱估计技术

摘要

The spatial spectrum estimation is a hot issue of underwater acoustic array signal processing re-search.It is generally believed that the information of target and background interference is contained in the co-variance matrix of received signal,and thus most of the existing spectral estimation techniques study the proper-ty of the covariance matrix of the received signal.Because of the correlation between signal and noise,it will maximize the noise component of the covariance matrix primary diagonal elements.Aiming at this feature,the underwater acoustic array sampling matrix inversion (SMI)minimum variance distortionless response (MVDR) beamforming technique is proposed based on diagonal reduction,on how diagonal reduction affects output signal is deduced to noise ratio in the output power spectrum target direction from a theoretical perspective,and the se-lective principle of the optimal diagonal reduction coefficients is analyzed.And according to the actual engineer-ing application,how to obtain the optimal diagonal reduction coefficients from the finite snapshot estimation sampling matrix is studied.This method does not need to estimate the number of signal sources.By algorithm simulation and sea trial data processing,through comparisons,based on diagonal reduction,validates the validi-ty and reliability of SMI-MVDR spatial spectral estimation.In the underwater acoustic array signal processing environment whose background noise level is strong,the proposed method can effectively improve the ability of the sonar multi-target resolution under Gauss white noise background.%空间谱估计一直都是水声阵列信号处理研究的热点问题，而一般认为目标和背景干扰的信息都蕴含在接收信号协方差矩阵中，因而现有的谱估计技术大多针对接收信号的协方差矩阵进行处理。由于信号和噪声的相关性，导致协方差矩阵中主对角线元素上的噪声分量最大。针对这一特性提出了基于对角减载的水声阵列采样矩阵求逆（sample matrix inversion，SMI）最小方差无畸变响应（minimum variance distortionless response， MVDR）波束形成技术。理论推导了对角减载量对输出功率谱目标方位上输出信噪比的影响，分析了最佳对角减载系数的选取原则。并针对实际工程应用情况，研究了由有限次快拍数估计的采样矩阵得到最佳对角减载系数的方法，该方法无需预估信号源数。通过算法仿真和海试数据处理，对比验证了基于对角减载的SMI-MVDR 空间谱估计的有效性和可靠性。对于背景噪声级较强的水声阵列信号处理环境来说，所提方法能有效提高高斯白噪声背景下的声纳多目标分辨性能。