Radio signal modulation recognition

The brief

Spectrum monitoring needs to know what kind of signal is on the air. Train a deep-learning model that recognises the modulation (BPSK, QPSK, QAM…) from raw I/Q samples, and test how it holds up as noise rises. Good looks like: accuracy reported against signal-to-noise ratio, and a clear comparison with a feature-based method.

What you will build

1. Load and visualise I/Q samples per modulation
2. Build a feature-based baseline
3. Train a CNN on raw samples
4. Plot accuracy against SNR
5. Explain where the classes get confused

Free, open data to build it with

• RadioML datasets ↗ DeepSig

Check each dataset's licence before you publish your work. Your mentor gives your team its own version of the brief (data, scope or region), so every team's project is its own.