This proposal describes the development of an analog-to-information (A2I) system capable of detecting and classifying non-periodic radio frequency (RF) signals such as noise radars, ultra wide band (UWB) radars, and low probability of detection/intercept (LPD/LPI) waveforms. The proposed system leverages advanced algorithms for non-periodic signal classification that were originally developed for marine mammal classification, and improves performance through the use of deep neural networks (DNNs). Phase-I is structured into three sub-phases: the Baseline Phase, the Mark-1 Phase, and the Mark-2 Phase . Preliminary results obtained by applying a developmental algorithm to a real-world data set containing non-periodic signals suggests that the proposed approach is capable of achieving state-of-the-art performance as measured by validation accuracy.
Benefit: A real-time system capable of consistently and accurately detecting and classifying non-periodic signals-of-interest has immediate applications in signals intelligence (SIGINT), electronic intelligence (ELINT), radar warning receivers (RWRs), missile warning receivers, and a number of other defense technologies. Such a system would also have immediate application to a number of biological and environmental monitoring problems like marine mammal population impact studies and seismic monitoring.
Keywords: ELINT, ELINT, Wavelets, SIGINT, deep neural networks, low probability of intercept (LPI), analog-to-information (A2I)