Radar Emitter Identification Without Mission Data File (MDF) by Using Voting Classifier & Multi-Layer Perceptron Hybrid Model to Improve Cognitive Electronic Support Measure (ESM) Capability

Abstract

Achieving air superiority is decisive in modern warfare, requiring fighter aircraft to detect and identify radar threats in contested electromagnetic environments. Traditionally, radar identification relies on matching parameters received by Radar Warning Receivers (RWR) with data stored in Mission Data Files (MDF) derived from Electronic Warfare (EW) threat libraries. However, missing entries or unknown radar modes can cause matching failures, leaving aircraft vulnerable. This study proposes a radar emitter identification method based on a hybrid Deep Learning and Machine Learning architecture. By leveraging threat library parametric data, the proposed approach eliminates manual MDF preparation and reduces dependency on exact parameter matching across ESM and Signals Intelligence systems. Comparative analyses show that the Voting-MLP hybrid model achieves an identification accuracy of 99.71 % and a classification accuracy of 99.75 % on synthetic radar datasets, correctly identifying 76 of 79 emitters. The improved strategy successfully predicts all 79 radar identities at a 50 % threshold, achieving the highest accuracy while maintaining operational reliability.