Radar-Specific Emitter Identification With Only Envelope Power Based on Multidimensional Complex Noncentral Chi-Square Classifier

The signal-level emitter individual identification can provide more accurate and abundant information for radar reconnaissance, interpretation, and combat capability judgment. Unintentional modulation on pulse (UMOP) caused by the differences in the transmitter hardware is distinctive to each emitter. Thus, it is believed as a potential approach to realizing specific emitter identification (SEI) for radar. However, the approach would still meet many issues, like the UMOP is subtle and covered by noise. Meanwhile, such a “fingerprint” exhibit an irregular fluctuation pattern, so it is hard to build a standard and unified feature extraction procedure. In this article, to avoid the influence of pulse modulation on the fingerprint features, we only use the fluctuation of the envelope power to realize SEI. In particular, a probability derivation is demonstrated, and it is concluded that the normalized power of UMOP obeys the complex noncentral chi-square (CNCX) distribution, which establishes a theoretical basis for SEI through UMOP in power. Based on it, a high-precision reconstruction of the UMOP template through several noisy samples is accomplished. Afterward, a Bayesian classifier is naturally built based on the reconstructed UMOP template, which is highly explainable and retains almost the complete distinguishing information of individuals in power. Simulation experiments show that the Bayesian classifier achieves a satisfying accuracy on radar SEI under a low signal-to-noise ratio (SNR) environment. Also, it has a significant advantage over the deep neural network (DNN) under both conditions where sufficient and few-shot training samples are assumed.