最小方差无失真响应波束形成解卷积后处理算法

为提高最小方差无失真响应（Minimum Variance Distortionless Response，MVDR）波束形成算法的方位分辨能力，本文将MVDR算法的输出功率谱重新建模为卷积的形式，并运用两种解卷积技术对MVDR的方位谱进行后处理。该算法将角度空间中心位置的单个声源的MVDR方位谱当作点扩散函数（Point Spreading Function，PSF），并运用Richardson-Lucy算法和快速迭代收缩阈值算法（Fast Iterative Shrinkage-Thresholding Algorithm，FISTA）分别对MVDR（MVDR-RL，MVDR-FISTA）的方位谱进行解卷积后处理，获得背景级更低的MVDR-RL和MVDR-FISTA方位谱，同时提高了分辨能力和估计精度。仿真实验显示了所提算法的良好性能。 

Deconvolution Post-processing Algorithm of Minimum Variance Distortionless Response Beamformer

In order to improve the azimuth estimation capability of the minimum variance distortionless response （MVDR） beamforming algorithm， the output power spectrum of the MVDR algorithm was reformed as a convolution form， two deconvolution post-processing techniques were proposed to deconvolve the azimuth spectrum of MVDR. The algorithms take the MVDR azimuth spectrum of a single sound source at the center of the angular space as a point spreading function （PSF）， and use the Richardson-Lucy algorithm and the fast iterative shrinkage-thresholding algorithm （FISTA） to deconvolute the MVDR （MVDR-RL， MVDR-FISTA） azimuth spectrum. The MVDR-RL and MVDR-FISTA azimuth spectrum has lower background level， higher resolution， and higher estimation accuracy. Simulation experiments show the good performance of the proposed algorithm.