Joint Estimation of Azimuth and Elevation via Manifold Separation for Arbitrary Array Structures

Direction-of-arrival estimation is a ubiquitous task in array processing. The conventional MUltiple Signal Classification (MUSIC) method is often computationally expensive, particularly in the application of joint azimuth and elevation estimation. In this paper, we propose an efficient way to compute the two-dimentional (2D) spatial spectrum. By using the manifold separation technique, we find that the 2-D null-spectrum has the form of discrete Fourier transform (DFT). Moreover, the coefficients of the DFT form can he truncated and computed in parallel. Then by exploiting the fact that the nonzero coefficients are concentrated only at the top-left corner, we present a partial 2-D DFT to compute the 2-D spatial spectrum, which can reduce the latency significantly. In order to improve the estimation performance, we formulate an iterative multidimensional-subspace-fitting estimator in which we apply the first-order Taylor expansion to expand the projection operator. At each iteration, the estimation errors can be computed by solving a set of linear equations. Numerical results demonstrate that the proposed estimator offers better estimation performance compared with the classical 2-D MUSIC.