An Open Channel Identifier using GNU Radio
summary
Claim: a five-layer feed-forward neural network implemented with GNU Radio 3.8 flowgraphs and SGD distinguishes open vs occupied Wi‑Fi channels more accurately than a classical power estimator but costs more compute. Built 800 simulated IQ files (split 640 train / 160 validation), sample rate 166 MS/s, SNR 44 dBm, converted to magnitudes before training. Results report 66.17% vs 49.12% accuracy, analysis shows random frequency selection is near-uniform and the NN is higher complexity; code and data are on the linked GitHub.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:23Spectrum Bullpen motivation
- 3:12Machine learning in radio
- 8:15Channel identifier algorithms
- 10:42Software setup and data
- 13:11Power estimation algorithm
- 15:22Neural network implementation
- 17:35Model performance analysis
- 19:42Data randomness and complexity
- 22:10Results and future work
- 23:39Current projects and conclusion
related talks
23mDeep learning inference in GNU Radio with ONNX
32mMan or Machine?: Developing a "Turing Test" for Radio Intelligence
16mCloud-LoRa: Enabling Cloud Radio Access LoRa Networks Using Reinforcement Learning...
17mSLNet: A Spectrogram Learning Neural Network for Deep Wireless Sensing
14mBrain-on-Switch: Towards Advanced Intelligent Network Data Plane via NN-Driven Traffic...
06mTeaching RF to Sense without RF Training Measurements
12mLightning Talk: Uplink Interference Optimizer, How to Optimize a Cellular Network...
56mChannel Gating for Cheaper and More Accurate Neural Nets