Solid State Radar Emitter Identification

The ultimate goal of this SBIR effort is to deliver improvements to existing Fleet electronic warfare support systems by using statistical learning techniques to aid in the classification and unique identification of modern radar emitters. Since software-defined and solid-state emitters do not have the analog features traditionally used for classification, we will use more abstract, but well-defined, feature extraction techniques in the joint time-frequency domain to exploit information hidden inside the noisy Radio Frequency (RF) signal that can be used to differentiate between multiple emitters.SEA CORP has conducted a proof-of-concept demonstration of the feature extraction, feature selection, and classification processes using real-world, labeled datasets. Processes and scripts were created to import Burst Digitized Intermediate Frequency (BDIF) and Continuous Digital Intermediate Frequency (CDIF) data into manageable-sized data structures. Initial classification results from 5,000-instance subsets created at random from a database of more than 550,000 unique instances showed up to 91% correct classification performance at the emitter level. Continued optimization of the feature extraction and feature selection algorithms will allow testing on larger datasets with the ultimate goal of transitioning our Solid State Radar Identification (SSRID) process into Navy Program of Record systems.