Robust Adaptive Beamforming In The Presence Of Strong Desired Signal

Robust adaptive beamforming has been one of the research focuses of the array signal processing, since it remains good performance in the presence of various steering vector mismatches. However, most of existing robust adaptive beamforming algorithms are only applicable to the cases where the desired-signal-free data are available or the power of desired signal is weak. The performances of them degrade dramatically if the desired signal becomes strong enough. Few studies have been done so far on the situation that the desired signal with a large range of signal-to-noise ratio (SNR) is present in the training data with imprecise direction-of-arrival (DOA) knowledge. Therefore, this issue of research is still in its infancy. The focus of this thesis is on the performances of robust adaptive beamforming algorithms in the presence of DOA mismatch and desired signal with a large range of SNR.This thesis starts from the standard Capon beamformer, which has high resolution and excellent interference suppression ability when the power of desired signal is weak and the steering vector is known accurately. The explicit expression is introduced that how the performance is affected by SNR. Simulation results demonstrate that the Capon beamformer treats the desired signal as interference and the performance is decreased inevitably in the presence of strong desired signal and DOA mismatch. In the past decades, various adaptive beamformers have been proposed to improve the robustness against steering vector mismatches in the weak desired signal situations, but they will fail when the desired signal becomes strong enough and the DOA mismatch exists.Since the general beamformers cannot be applied to the situations this thesis focuses on, some novel robust adaptive beamforming algorithms proposed in the recent years are introduced. Those algorithms try to remove the desired signal effects and keep robust against the DOA mismatch. In particular, owing to the elimination of the desired signal component from the covariance matrix, the robust adaptive beamformer based on interference-plus-noise covariance matrix reconstruction enjoys quite good performance over a large range of SNR. However, this beamformer does not have unit response in the actual DOA of the desired signal due to the imprecise DOA knowledge. Therefore, beamformers based on covariance matrix reconstruction and three different steering vector estimation methods are introduced to improve the performance. Simulations are carried out to compare the algorithms mentioned above, which show that the novel beamformers outperform the general beamformers when the desired signal becomes strong enough and the DOA mismatch exists.Nevertheless, it’s difficult to apply the above mentioned algorithms to the practical engineering. And the least-mean-square (LMS) iterative algorithm is still the most commonly used in the practical engineering. In the end of this thesis, an iterative algorithm combined the LMS algorithm and interference-plus-noise covariance matrix reconstruction technique is developed. Simulations demonstrate that the performance of the convergent iterative algorithm is close to the optimal values over a large range of SNR.