Classifier of over-the-air recordings of 24 digital and analog modulation of radio signals based on deep learning ResNet architecture.
A dataset that includes both synthetic simulated channel effects and over-the-air recordings of 24 digital and analog modulation types has been heavily validated.
This dataset was used for Over-the-air deep learning-based radio signal classification published 2017 in IEEE Journal of Selected Topics in Signal Processing, which provides additional details and a description of the dataset.
Data was stored in hdf5 format as complex floating-point values, with 2 million examples, each 1024 samples long. I converted the data to .npy format since I found it took much less time to load. You can find that below: Link
I conduct an in depth study on the performance of deep learning based radio signal classification for radio communications signal.
I was working with one third part of dataset (851968 signals), I found myself running into problems with memory shortages on Kaggle’s default and decided to use powerful Google Cloud Platform virtual machine instance. I used the virtual machine instance n1-standard-8 (8 vCPU, 30 GB memory) with a NVIDIA Tesla V100 GPU to run my training session. This required about 7.5 hours of training time
The model achieved maximum accuracy of 95.72% on the clean signal dataset: Signal-to-noise ratio > 8 dB.
Note that the maximum classification accuracy of 62% if you want to test the model to its absolute limits on a mix of clean signals and signals with very high interference. Some of the signals have so much noise that they are virtually unrecognizable.
Thanks to Tim O’Shea, Tamoghna Roy, T. Charles Clancy for posting Over the Air Deep Learning Based Signal Classification.