DOA Estimation Based On Compressive Sensing

The theories and methods for target DOA estimation have a great importance in military field. Spatial spectrum estimation based on the array is the main method of DOA estimaton since it achieves high resolution. However, this kind of methods requires a mass of samples, which is difficult to carry out on the war field, so that it is urgent to explore a novel method, which only needs far fewer samples for high-precision DOA estimation. The theory of Compressive Sensing (CS) is a novel sensing/sampling paradigm that goes against the common wisdom in data acquisition. CS theory asserts that one can recover certain signals or estimate their parameters from far fewer samples or measurements than traditional methods use when the signals are sparse or compressable. In a nutshell, CS theory acquires the bottleneck on Nyquist rate. This paper surveys the application of CS theory on DOA estimation.Combined with the mathematic model of DOA estimation, we research the DOA estimation method based on CS theory and the array sampling systems design using for target location. For the former, we exploit the sparsity of source signals by using uniform sine grid division. It is well-known that the sensing matrix satisfying the RIP or MIP can ensure the good performance of signal reconstruction within CS framework. Through the theoretical analysis and the simulations, we conclude that the array manifold matrix obtained by uniform sine grid division, which is correspond to the sensing matrix in CS theory, has better RIP and MIP than that obtained by uniform angle grid division used in other literature generally. And compared with the classical MUSIC algorithm used on DOA estimation, the presented method has many advantages, including sample reduction, no any prior knowledge of sources required, improved robustness to noise, and estimation DOA for large numbers of sources and so on. It is shows that the presented method based on CS with uniform sine grid division is a novel and excellent method for DOA estimation.For the design of array sampling system, based on the theory of CS with coherent and redundant dictionaries, we propose two kinds of design scheme which will offer the linear measurements of the senors receiving datas on the Gauss random matrix and the random sample matrix, respectively. The simulations shows that the performance of DOA estimation based on these new array sampling systems and CS reconstruction algorithm is more perfect than the former, say nothing of MUSIC algorithm, especially the performance od robust to noise and high resolution. Besides, the array sampling systems are quite easy to realize in the practice application. Spacially, the system correspond to random sample matrix is more remarkable both on the performerance of DOA estimation and the conveniency of system realization than that correspond to Gauss random matrix.The researches in this thesis provided basic theories and criterions in applications.