Research On Beamforming Algorithms

Beamforming is an important research topic in array signal processing and it has found many applications ranging from wireless communications,radar,sonar,and speech processing,to medical imaging,radio astronomy,etc.Among the art of adaptive beamforming algorithm,minimum-variance-distortionless-response(MVDR)beamforming receives abundant focus due to its good performance and flexible expression.Traditional MVDR beamformer however rely on assumptions of array calibration and signal model,while it suffers great performance degrading as those assumptions usually do not hold practically.Therefore,improving the robustness of MVDR beamformer is necessary.According to the weight function of MVDR beamformer,there are directly two approaches to improve MVDR beamformer.One is to estimate the interference-plus-noise covariance matrix in a better way;the other is to modify the steering vector of desired signal.Based on the two approaches of improving MVDR beamformer,this article proposed one new method for each approach.To reduce the computational complexity of current method of covariance matrix reconstruction,a new beamformer based on spatial power spectrum sampling is proposed in this thesis.The proposed method utilizes spatial power spectrum sampling and covariance matrix taper technology and can waive the spectrum estimation process in interference-plus-noise covariance matrix reconstruction,and thus greatly reduce the needed computational complexity.Simulation results have demonstrated that the proposed beamformer can achieve a very similar performance to its high-complexity version.To combat the vagueness of the feasible region in beamformers which are based on modifying the presumed steering vector,an intersection method has been proposed for more effective estimation of the S subspace,and further improves the robustness in beamforming.The new estimation is robust to steering vector mismatch and overestimation of the SI subspace,capable of detecting the relative strength of the desired signal.With these properties,the estimated S subspace can be used to reduce the steering vector mismatch error of the desired signal,which in turn leads to improved beamforming performance for several representative robust adaptive beamforming methods,as verified by simulation results.