Two Dimensional Hybrid MUSIC Method Applied To Arbitrary Arrays

Array signal processing is an important branch of signal processing. It is widely used in daily life and scientific research in many fields. Direction-of-arrival (DOA) estimation is a ubiquitous task concerned in array processing. Its purpose is to determine the direction of arrival of a plurality of target signals in a region of space. The well-known Multiple Signal Classification (MUSIC) method, a classical subspace type method, asymptotically exhibits infinite resolution capabilities and thus are classified as a super-resolution technique in estimating DOA. The traditional search-based MUSIC method is often computationally expensive, particularly for the application of joint azimuth and elevation estimation.In order to alleviate the computational burden, the root-MUSIC algorithm can convert the DOA estimation to a problem of polynomial rooting by exploiting the Vandermonde structured array manifold vector of uniform linear arrays (ULA). It can reduce the computation burden to some extent. By using the manifold separation technique (MST),the conventional root-MUSIC can be extended to arrays with arbitrary geometry while the 2-D DOA estimation can be converted to a problem of bivariate polynomial rooting. However, if the array aperture becomes large, the order of the the bivariate polynomial must be quite large which leads to unacceptable computational complexity for the rooting procedure. But a computationally complex polynomial rooting procedure is still required. Therefore, It is an important research direction to study more joint hybrid algorithms for the two-dimensional DOA estimation. These algorithms should be applied to more array structure, a smaller amount of calculation, the calculation process more concise and more robust.Firstly, the manifold vector of nonuniform linear array is modeled by using manifold separation technique. Then we obtain a matrix array of samples by using the method we mentioned. To obtain noise eigenvectors and signal eigenvectors, we make the eigen-decomposition of the sampling matrix by using Jacobi rotation.In this thesis, we propose a computation attractive 2-D direction-of-arrival (DOA) estimator which can be viewed as a hybrid MUSIC-based method. We use the MST method to convert the 2D-MUSIC cost function into stand 2D-FFT form. In doing so, the 2-D spatial spectrum can be obtained by using 2D-FFT. Since a relatively small point number for the FFT is chosen, the DOAs are located roughly. Then the 2D-MUSIC method with fine angular grid is utilized to search the DOAs finely within a small angular section.The proposed hybrid method not only alleviates the computation burden of root-MUSIC or MUSIC solely used; it also achieves almost the same DOA estimation performance and is easy to implement.