Classification of Radar Targets using Features Based on Warped Discrete Fourier Transform

In this research the signals from ground surveillance radar audio output are classified using hierarchical nonlinear classifier based on Support Vector Machine. Central Doppler frequency and spectral width around it are used as features. These features are obtained based on spectrogram calculated using discrete Fourier transform and spectrogram calculated using warped discrete Fourier transform. The performance of used features for classification are analyzed through classification accuracy. Obtained results show that features extracted from the spectrogram calculated using warped discrete Fourier transform provide higher classification accuracy.

I. Introduction

The situational awareness on the battlefield is one of the main task of sensors application for military purposes [1]. The complexity of the battlefield implies usage of sensors that work in the different part of the electromagnetic spectrum: radar, visible cameras, thermal cameras, acoustic sensors, etc. Each sensor utilization has its advantages and shortcomings. Comparing to radars, visible and thermal cameras provide higher resolution and they are passive sensors. On the other hand, the possibility of working in complex atmospheric conditions (rain, fog, snow, etc.) and the higher detection range, are advantages of radar sensors.