Sound Beam Forming And Blind Signal Separation

Beamforming is an important task in array signal processing with applications, among others, in radar, sonar, acoustics, astronomy, seismology, communications, and medical-imaging. Under ideal conditions,the adaptive beamformer has excellent performance. It has good interference rejection capability and maximizes the signal-to-interference-plus-noise ratio (SINR). But in practical scenarios, the performance degradation of traditional adaptive beamforming techniques may become even more pronounced, because most of these techniques are based on the assumption of an accurate knowledge of the array response to desired signal. Robust adaptive beamforming becomes an important research topic. Algorithms for blind source separation aim to select a desired source while suppressing interfering sources, without specific knowledge of the sources or channel. Because many communication signals have a constant mudulus proterty and all information is carried in the phase, the constant mudolus algorithm is a research focus of the blind source separation problem. Robust adaptive beamforming and constant modulus algorithm are reasearched in the text.(1) The paper reviews systematically some conventional beamforming algorithms. The performance of MVDR is analyzed when the look direction errors exist. And some classic beamforming algorithms are simulated.(2) Four robust adaptive beamforming algorithms are studied,which are the norm constrained Capon beamformer (NCCB), robust Capon beamformer (RCB), doubly constrained robust Capon beamformer (DCRCB) and robust minimum variance beamforming based on worst case performance optimization. NCCB algorithm offers a good performance in output SINR, but it is not clear hoe to choose the constraint value based on information about the uncertainty of the array steering vector. We coupled the covariance fitting formulation of SCB with an ellipsoidal uncertainty set to obtain a RCB and DCRCB. Their computation complexity are close to SCB's.Rank-one signal case and general-rank signal case are researched in the algorithms of robust minimum variance beamforming based on worst case performance optimization. And they achieve the SINRs that are close to the optimal one.The least-squares constant modulus algorithm (LS-CMA) and algebraic constant modulus algorithm (ACMA) are studied. Based on MMSE criterion, the LS-CMA obtain weight vector by the stochastic gradient method, and it has fast convergence speed. ACMA uses the constant modulus property, eigenvalue and joint diagonalization in order to separate signals. Simulation results verify the validity and feasibility of the two algorithms.