Researches On Robust Adaptive Beamforming

Robust adaptive beamforming is an important issue for the array signal processing.The existing robust beamformers mainly consist of eigenspace-based beamformers,worst-case-based beamformers, covariance-matrix-reconstruction-based beamformers,intelligent algorithm-based beamformers, and diagonal loading beamformers.Diagonal loading beamformers with stable performance, simple structure, and lowcomputational burden are widely used. The key point of the algorithm is to calculate theloading factor. Two algorithm of calculating the loading factor are proposed in thisthesis, diagonal loading based on Gaussian distribution, and diagonal loading based onprobabilistic constraint. The diagonal loading based on Gaussian distribution is theimproved GLC-based diagonal loading beamformer, which reduce the amount of thecalculation effectively with the performance undegraded. The diagonal loading based onprobabilistic constraint remains the advantages of the diagonal loading algorithm andthe probabilistic constraint based on worst-case optimization, and determines theloading factor via the probabilistic constraint. The algorithm needs a parameter to be set,to which the algorithm is robust as shown in the simulation. So the parameter can befixed (such as0.2), and the algorithm can be seen as parameter-free algorithm. Thesimulations show its better performance compare with existing diagonal loadingbeamformer and eigenspace-based beamformer.With small snapshots, the estimation error of the covariance matrix will lead thegreat degradation when using standard Canpon beamformer. There has been someanalysis showing that the exchange of signal space and the noise space in the covariancematrix with small snapshots is the major factor of the degradation. In this thesis, anintensive study on the factor of the degradation is made by proposing a model ofcross-correlation. The model explains the principle of the existing robust beamformersuccessfully, and the theoretic performances based on this model come the same withthe result of the existing references.A robust beamformer based on the parameterized estimation of the covariancematrix, which removed the cross-correlation item in the estimated covariance matrix effectively. The algorithm takes full advantage of the prior information that the signalsof the space array are sparse and low-dimension, and gets a better estimation of thecovariance matrix. The simulations show its excellent performance approximating theanalytic optimal one. And the computational burden is much lighter than the one basedon the covariance matrix reconstruction with the same performance, and even lighterthan the diagonal loading beamformer.