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 minimum 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