Some Researches On Direction-of-Arrival Estimation

With the development of communication technology, communication signals such as short-time and burst signals have been widely used, and this brings forward more strict demand for the real-time ability and validity of direction finding(DF) system. This paper studies several problems in direction-of-arrival(DOA) estimation, especially the DF algorithms which has better real-time property and validity, and the selected algorithm has been implemented on chip ADSP-TS201. Meanwhile, the affects of array structure to performance of DF system are also discussed. The main content is as follow:1. The 2D-DOA estimation algorithms based on specified array structure are studied. Fast DOA estimation algorithms based on parallel linear and L-shaped array are mainly discussed. The problems such as parameters pair-matching and angle ambiguity of these kinds of algorithms are analyzed and the solution is proposed. Based on L-shaped array, an improved correlation matrix based ESPRIT algorithm is proposed, which can resolve mismatch problem under some special conditions. And then, the joint-SVD algorithm is improved to resolve angle ambiguity problem of original algorithm, the estimation performance in special directions is also improved.2. The 1D-DOA estimation algorithms based on arbitrary array are studied. Firstly, in order to avoid spectrum peak searching, two methods which convert spectrum peak searching into polynomial rooting are discussed: Manifold Separation Technology(MST) and Fourier domain MUSIC algorithm. Simulation and comparison have been taken, and the results show good performance. Secondly, by applying the two methods in parallel linear and L-shaped array, DOA estimation without 2D searching is achieved while the array is nonuniform.3. The 2D-DOA estimation algorithms based on arbitrary array are studied. Firstly, a new DOA estimation algorithm with auxiliary sensor which can be applied in arbitrary plannar array is proposed. The new algorithm uses spatial-time DOA matrix to construct rotation invariance by introducing an auxiliary sensor, thereby the elevation could be estimated first and then the azimuth could be obtained, therefore 2D-DOA could be estimated fast. Secondly, in view of huge computational cost in spectrum peak searching of traditional MUSIC algorithm, by combining Metropolis-Hastings sampler of Markov Chain Monte Carlo(MCMC) and sequential searching by eliminating correlativity, a fast 2D-DOA estimation algorithm based on arbitrary array is proposed.4. The relationship between array manifold and the problem of DOA estimation precision and ambiguity is studied. Firstly, the relationship between differential geometry parameters of array manifold and DOA estimation precision is investigated by using the tool of differential geometry. On the basis of this, approaches to optimize the performance of direction finding in linear and circular array have been put forward under the condition that aperture and sensors'number of array are fixed while one signal is imposed and the approaches are validated by simulation. Secondly, manifold ambiguities are analyzed with the tool of differential geometry, methods to calculate two types of"ambiguous generator sets"(AGS) are introduced. At last, two types of approaches for ambiguity resolution are proposed under identifiable conditions.5. Based on ADSP-TS201, the fast MUSIC algorithm via Metropolis-Hastings sampler and sequential searching has been coded. A series of measures have been taken to optimize the program and the data test results are given.