A New Electromagnetic Target Classification Method with MUSIC Algorithm

This paper introduces a novel method for aspect invariant electromagnetic target recognition based on the use of multiple signal classification (MUSIC) algorithm to extract late-time resonant target features from the ultra-wideband scattered data. This method is mainly based on the usage of MUSIC spectra obtained from electromagnetic scattered data as the target features. This approach achieves very high accuracy rates even at very low signal-to-noise ratio (SNR) values although it needs scattered data for classifier design at only a few different aspects and makes use of the MUSIC algorithm in a simple and computationally efficient way. Details of the theoretical background, the classifier design process and the results for conducting and dielectric spherical targets will be presented in the following sections.