基于交叉验证的MVDR波束形成对角线加载因子选择

A cross-validation method for selecting diagonal-loading factors for MVDR beamforming

最小方差无失真响应(MVDR:Minimum variance distortionless response)波束形成器在阵列信号处理中具有广泛的应用。在样本数少的情况下,MVDR波束形成器设计中的协方差矩阵的估计是一个挑战。已经证明对角加载(DL)是解决该问题的有效方法,但是对角加载因子(DLF:Diagonal loading factor)的选择是困难的。本文中,我们提出了一种基于留一交叉验证的(LOOCV:Leave-one-out cross-validation)数据驱动方法,用于为MVDR波束形成器选择DLF。我们还对算法做了简化使得DLF的选择可以在低复杂度下实现。数值结果表明,该方法可以实现DLF的近乎最优的选择。

Minimum variance distortionless response(MVDR)beamformers have extensive applications in array signal processing.The design of MVDR beamformers requires the estimation of covariance matrices,which is challenging in low sample support scenarios.Diagonal loading(DL)has proven to be an effective approach based on LOOCV to address this issue but the choice of the diagonal loading factor(DLF)is difficult.In this letter,we propose a data-driven approach for choosing the DLF for MVDR beamformers.We show that the choice of the DLF can be achieved at low complexity.Numerical results suggest that the proposed approach can achieve near-oracle choice of the DLF.