Recognition performance analysis of instantaneous phase and its transformed features for radar emitter identification

Abstract

Instantaneous phase (IP) feature is a time-varying curve describing the unintentional modulation on the phase. IP is a basic feature for identifying different radar emitters, and mostly other features can always be obtained from its transformations. Thus an important issue is to choose a suitable transformation of IP which can achieve the best recognition performance, more specifically, the highest recognition ratio. Heretofore, the recognition ratios of transformed IP features were derived basically through experiments, lacking theoretical calculation and analysis. In this study, the authors first extract the IP feature from a given radar pulse, and then propose a unified model for various transformations of IP. After that, they calculate the recognition ratio theoretically based on this model. Subsequently, they prove that any linear transformation, such as derivative, smoothing, and linear dimensionality reduction cannot improve the recognition performance compared to the original IP. They also find the essentially mathematical reasons of the seemingly unimaginable results. Finally, they validate the analysis results through both computer simulations and experiments on measured data collected from 120 radar emitters in the real environment. This work provides a theoretical basis for feature extraction and feature assessment in radar emitter identification.