Beamspace Transformation Matrix Design Using Numerical Optimization

This paper addresses the problem of designing a data-dimension reducing transformation matrix, to be used for beamspace direction-of-arrival (DOA) estimation. The design of the transformation matrix is based on numerical optimization techniques. The criteria to be satissed is to retain as much as possible of the achievable, optimal estimation accuracy using the non-reduced data-sizes, while also taking into account the sidelobe levels of the beampattern. Comparisons of estimation accuracy and sensitivity to out-of-sector emitters for the design methods considered herein are carried out by means of simulation examples, using the WSF-method for the DOA-estimation. In order to reduce the number of parameters to be optimized, a parametrization of the transformation matrix is made, that utilizes the properties of Householder-reeections.