Our vision pairs a convolutional network that learns the waveform with the super-resolution direction finding this team has practiced for decades. The network names the threat. The array puts a range and bearing on it. Together they keep a crowded band usable.
In-phase and quadrature samples become an image of the band. A compact convolutional network — MobileNet-class models are the portability target — learns waveform fingerprints and returns class, confidence, and frequency.
Classification key performance indicators point super-resolution direction finding, including MUSIC-class angle-of-arrival, at the energy that matters. Two or more separated arrays support triangulation.
Class plus direction, ready for fusion, command and control, and a tracking radar. The effector engages on that cue.
The sensor rides commercial-derived software-defined radios and coherent arrays — AD9361-class receivers and multi-element IQ paths — so the capability arrives with the hardware already in reach. The same classifier maps onto a semi-custom FPGA or a dedicated edge engine, with compute about the size and cost of the radio.
Today’s lower-priced RF and layered kits still sit in a capital band, often in excess of $100k. Mind & Signal is building the passive sensor about a tenth of that price — so a convoy, and then a formation, can carry one per platform.