Performance Analysis for Finite Sample MVDR Beamformer With Forward Backward Processing
The employment of forward-backward (FB) processing to both batch and adaptive beamformers provides a superior performance than the forward only (FO) algorithm. Based on the finite-sample min- imum variance distortionless response (MVDR) beamformer, a set of closed-form expressions for the variance of the weight vector, the output signal-to-interference-plus-noise ratio (SINR) and the mean-square error (MSE) is derived, for both FB and FO algorithms. As shown in our analysis, all errors of the FB algorithm are only approximately half of the FO case. Index Terms—Finite sample, forward backward, MSE, MVDR beam- former, SINR, Wishart distribution.