SIFTER – Signals intelligence field tool for enhanced recognition

Military platforms face an increasingly cluttered radio frequency (RF) environment. There is particular need for continuous, real-time automated monitoring, which would benefit from detecting the signals from unmanned aerial systems (UAS) and ground surveillance radar in an expedient way. Tracking numerous frequency spans at once could be automated through the technology of radio frequency machine learning (RFML). Most development activity in RFML is limited to automatic modulation classification (AMC), which is only the determination of a signal’s modulation type (AM/FM/Phase Shift/etc.). Yet this is not sufficient to identify the specific type of transmitter. Furthermore, most RFML is focused on a locked-in number of specific signal type classes. In contrast, we are developing an innovative RFML clustering system that can group an open set of never-before-seen signals, and then match new samples of those signals to the appropriate group automatically. We propose to extend and adapt these RFML tools to an embedded architecture that can take advantage of the very high digitization rates afforded by modern software-defined radio receivers, with the aid of a priority attention algorithm.