Signal Discovery: rapidly find signals of interest in large datasets

RFeye DeepView is a forensic signal analysis software for advanced signal monitoring and measurement. Users can reliably record and capture RF signals (I/Q data), which can be analyzed and further processed. After recording terabytes of I/Q data, the new Signal Discovery feature in RFeye DeepView enables users to glean specific information from a large amount of data.

As monitoring and intercepting signals—that often have wide bandwidth and highly complex, low-power characteristics—across wide frequency ranges can be very challenging, signal analysis of unknown and complex emitters is problematic.

However, the Signal Discovery feature in RFeye DeepView allows users to efficiently carry out statistical analysis of large datasets—allowing them to rapidly identify anomalies that can lead to discovering signals of interest.

How does Signal Discovery work?

Modern communication revolves around pulses and power. Signal Discovery gives EW and SIGINT operators and test engineers who want to identify, and ultimately geolocate, an unknown transmission the ability to understand pulse parameters more clearly.

image (1)

Image 1: When hovering over the signal, Signal Discovery shows the pulse descriptor.

Recording I/Q data over a long duration creates a database of pulse descriptors; however, this results in a sea of analog information. Signal Discovery allows operators to view all this information and make sense of it statistically by searching for hotspots.

By querying the database for a particular pulse type in real-time, users can quickly isolate those specific pulse types from the rest of the recording—which is not realistic to undertake manually.

Essentially, Signal Discovery allows users to identify statistical anomalies by discriminating between what they expect and what they do not expect. They can then quickly zoom into patterns of signals to reduce the number of signals from over a million to one hundred.  

Example: filtering signals in a crowded Wi-Fi band (2.4 GHz)

With a statistical view, users can see all signals in the spectrum.

image (2)

Image 2: The flexible view in Signal Discovery displays the frequency, time, power, bandwidth, pulse duration, and pulse repetition rate of the signals.

The top right graph in Image 2 shows signal hotspots that are combinations of pulse bandwidth and pulse duration. However, no human operator can analyze 1.4 million signals.

Signal Discovery allows the user to select a hotspot area of interest, rapidly reducing the number of signals to 51,722.

image (3)

Image 3: Selecting a hotspot area reduces the number of signals to 51,722.

Identifying an unexpected statistical hotspot allows the user to see patterns. By continuously zooming into hotspots of interest (which isolates signals of interest), users can reduce the number of signals and see statistically similar signals.

image (4)

Image 4: Further filtering reduces the number to 1449 signals.

Users might see statistical patterns based on time of day or specific days—patterns that are unlikely to have been noticed without the ability to zoom into areas of interest.

Image 6&7

Image 5: Further filtering of signals in the dataset

The images above show how confining the view to certain statistical parameters further reduces the number of signals from 103 to 16. 

Once the user has a manageable number of signals, further analysis can be carried out of these signals to identify a particular transmitter. While the Signal Discovery feature does not tell the user the nature of these signals, we statistically expect to see certain types of signals when looking in a specific band.

After using Signal Discovery to identify LPI signals of interest, users can use DeepView to analyze those signals further.

If a SIGINT or EW operator or test engineer identifies an enemy signal, they can build a detector against that signal. When loaded into real-time spectrum monitoring software RFeye Site, detectors look at a signal’s frequency, power, and time characteristics, which they compare to a library of signals. When they find a match, the detectors automatically trigger a workflow geolocation to geolocate the emitter.

This method is highly efficient in differentiating relevant signals from irrelevant ones in crowded and complex electromagnetic environments, making it ideal for real-time applications where accuracy and speed are critical.

Jaimie Brzezinski

Jaimie Brzezinski is Head of Content for CRFS. His specialty is turning highly technical ideas into engaging narratives. He has 15+ years of experience in writing technical content and building global teams of subject matter experts.

Get instant news & expert insights

Join thousands of professionals who subscribe for exclusive insights and early access to our reports and analysis.

Subscribe now