RF Signal Intelligence

Automatic
Modulation
Recognition

Deep learning that adapts to your hardware — high-accuracy signal classification across contested RF environments.

Request a Demo Learn More

Neural Architecture

I(t) Q(t) BPSK QPSK 2FSK 16QAM ▸ INPUT OUTPUT

Edge-deployable RF classification
without hardware lock-in.

01

Modulation Recognition

Classifies BPSK, QPSK, 2FSK, and 16QAM signals in real time. Trained on synthetic IQ data spanning a wide SNR range, with no reliance on proprietary datasets or hardware signatures.

02

Hardware-Adaptive Intelligence

The model learns the RF characteristics of your front-end and self-optimizes — no manual recalibration. Deploy against USRP, PlutoSDR, RTL-SDR, or recorded IQ captures and the classifier adapts to what it sees.

03

Low-Overhead Architecture

A 3-layer convolutional network with 233K parameters. Designed to run on constrained platforms — no GPU required for inference, suitable for embedded and tactical deployments.

From synthetic data
to real hardware.

Models train on synthetic IQ data, then adapt in-situ to the RF characteristics of whatever front-end they're deployed on — learning hardware-specific noise profiles, gain curves, and frequency response without manual recalibration.

The result: a classifier that gets sharper the longer it runs on your hardware, not one that needs to be retrained every time you swap a front-end.

USRP B210 software-defined radio

Presenting at MILCOM 2026

We are seeking to present and demonstrate our AMR research at IEEE MILCOM 2026. If you are an organizer, program chair, or prospective collaborator, we'd welcome a conversation.

Get in Touch

Start a conversation.

Whether you're a conference organizer, a defense researcher, or exploring RF intelligence for your program — reach out and we'll get back to you within 48 hours.

hello@faradaylabs.xyz