Spectrum awareness is a watchword in these times of crowded airwaves, especially for 5G mobile network operators. Managing coexistence in the Citizens Broadband Radio Service (CBRS) band in the U.S. means increased vigilance through continuous spectrum monitoring as well as coordination between 4G/LTE and 5G New Radio (NR) cellular and military radar operators in shared bands.
In its OmniSIG software, DeepSig brings machine learning to bear on the growing requirements for RF situational awareness. The software takes input from any radio and performs real-time detection and classification of signals. Working from raw IQ data, the software’s neural network scans for signals, known and unknown, and labels them for operators’ informational purposes.
DeepSig’s tools can also perform some post-acquisition analysis of signals. The software detects and classifies signals in less than 10 ms. It also takes a wideband look at the spectrum, further ensuring that it will grab whatever is out there.
In the accompanying video, DeepSig’s director of operations, John Ryder, provides a demo of OmniSIG using a Tektronix real-time spectrum analyzer to feed RF input to the tool. Watch to see how the software quickly identifies and labels signals across the band.
Another benefit of OmniSIG is the ability to update the tool’s machine learning with real-world data. New signal-detection models can be developed in less than 24 hours, with either a signal-recording feature or by feeding pre-recorded RF signals to the tool. Such signals are labeled by operators, but the software learns that signal’s features for future recognition.
