UWB target classification using SVM

Abstract

Ultra-Wideband(UWB) radar signals are characterized for having both high frequency carrier and high bandwidth. This makes the scattered field from the targets when irradiated with UWB pulses highly dependent of the composition and shape of the target. Our goal is toclassify objects by their composition from their scattered responses. In this paper, we propose to use a Support Vector Machine (SVM) to solve the problem for distinct dielectric materials and sphere elements. For a problem considering Mdifferent materials andRradii, we compare performance of three different SVM configurations. The first one considersthe general problem where each class corresponds to adifferent material. In this approach, each class is trained withdata corresponding to all Rradii. On a second approach,we classify by both radii and material. This gives a largerproblem to solve, where the number of classes of the SVM is M×R+ 1. Finally, a third approach considers a cascadeof SVMs where the first layer consists of a SVM for R+ 1classes, each class associated with one radius, while the secondlayer is composed of Rdifferent SVMs, each corresponding to a different radius, that classify between the Mmaterials. Monte Carlo experiments are run to compare performanceamong the different proposed schemes. We analyze the results considering both classification and algorithmic complexity.