Reducing the computational complexity of FIR 2D fan and 3D cone filters

Abstract

Plane waves are an important class of multidimensional signals which are often en­countered in array signal processing applications such as, astronomy, radar/sonar, seismology, biomedical imaging, directional audio systems and wireless communica­tions. In many of these applications, it is of interest to selectively filter or enhance PWs on the basis of their space-time directions of arrival. A class of 2D/3D filters, commonly known as fan (velocity)/cone filters, has been extensively used in various fields of array signal processing applications to accomplish the above mentioned task. In this thesis two novel methods are proposed to reduce the computational complexity associated with the Finite Impulse Response (FIR) implementation of such filters. The first method takes into account the a priori knowledge of the dominant transients of fan/ cone impulse responses and the second method considers spatial undersampling as a means of complexity reduction. It is shown through simulation and numerical examples that both methods provide significant amount of computational savings.