Object Recognition With Natural Resonance Annihilation Using New N-Pulse Waveforms and Kernel Density Discrimination Measure

Radar object recognition based on aspect-independent unique natural resonance features has been widely investigated. There are two main challenges: first, robust techniques to extract resonance features, and second, a synthesis of discriminant signals to achieve automatic quantitative object discrimination. In literature, various time-domain-based techniques are explored, where the discriminant signals are synthesized to make the late-time energy zero, once the target response is convolved with the true pulse from the waveform library. This approach is required to have an accurate estimation of a late-time index, along with its main focus on annihilating resonance information. However, to achieve accurate object recognition, there is a need to have a nonzero or maximum response other than true discriminant pulses. Thus, to fulfill the two critical tasks, natural resonance annihilation and getting a nonzero response for nonresonant components, we proposed the synthesis of frequency-domain-modified notch-filter discriminant waveforms known as N-pulses. Moreover, we proposed the statistical quantitative discrimination measure of the kernel peak-density discrimination number (KDN). This article investigates N-pulses by considering the effect of polarization variations along with parametric analysis. Moreover, the simulation analysis for canonical dielectric-coated objects is conducted followed by the noise analysis.