Research On Super-resolution DOA Estimation And Robust Beamforming Techniques

Direction of arrival (DOA) estimation and adaptive beamforming are two important research contents in the field of modern array signal processing. They have had very wide applications in many fields, such as communications, radar, sonar, seismology and radio astronomy, etc. The basic model and theory of array signal processing are introduced in this dissertation firstly, then, the properties and disadvantages of the existed DOA estimation and beamforming algorithms are analyzed. Furthermore, the new algorithms are proposed and discussed.According to the study of the DOA estimation algorithm under the multipath environment, the concept of fading coefficient is introduced to overcome the disadvantage of array aperture loss caused by spatial smoothing algorithms. Then, an improved fading coefficient estimation method is presented. Moreover, a joint DOA and fading coefficient maximum likelihood estimation algorithm is studied. To reduce the computational complexity it applies an iterative technique called alternating maximization to simplify the multidimensional optimization problem to multiple one-dimensional procedures. Finally, the Cramer-Rao bounds for DOA and fading coefficient under the condition of known waveforms have been derived. It is noted that, several distinct signal environments in the multipath propagation are discussed in the simulation experiments of maximum likelihood algorithm. Simulation results show that the proposed maximum likelihood algorithm is well behaved in these situations.Three preprocessing techniques, which transform different frequencies into the same reference frequency to apply the narrowband method, are studied in the wideband DOA estimation algorithms. First, a wideband focusing algorithm based on the received array snapshots is proposed, that is, the solution of focusing matrices is independent on the array steering vector. The information of signal subspace is directly extracted from the received snapshots, thus, the steps of angle preprocessing are omitted to avoid the influence of the pre-estimated angle errors. Secondly, it is proved that the incident angles of wideband signals can be obtained by directly add up the covariance matrices in different frequency bins in theory, but its resolution is limited to the array manifold. Finally, a preprecessing method based on the frequency integration of array steering vector is presented. Although the performance of this method is not the best of all, its computational cost is very low and it is easy to hardware implementation. In comparison with the existed wideband algorithms, their advantages and disadvantages are pointed out to provide important references for practical engineering project.Robust algorithm is the effective method to overcome the uncertain errors of array steering vector and it is the key technique to realize in practice. Consider the computational complexity of the optimization algorithm under the ellipsoidal constraint and the difficulties of determining the parameters of ellipsoidal equation, we simplify the robust minimum variance algorithm based on the ellipsoidal constraint. The simplified algorithm overcomes the disadvantage of traditional minimum variance algorithm which is sensitive to the array steering vector errors when the errors are limited to some bounds. According to the derivation of robust beamforming algorithms with uncertain weights, it is found that these algorithms are independent on the signal steering vector. These algorithms, whose diagonal loading factor is close to the signal eigenvalues, are essentially a kind of negative diagonal loading method. Thereby the signal eigenvalue of sample covariance matrix approaches to zero. In comparison with the robust algorithms based on the uncertain array steering vector errors, the proposed algorithms do not need the estimated DOA of expected signal. These new methods not only reduce the computational complexity remarkably, but also prevent the phenomenon of signal cancellation effectively. However, if the nagtive loading factors are ill-chosen, these methods may fail due to the poor numerical stability.