AN OPEN-SOURCED TIME-FREQUENCY DOMAIN RF CLASSIFICATION FRAMEWORK

These datasets accompany the EUSIPCO 2021 paper: AN OPEN-SOURCED TIME-FREQUENCY
DOMAIN RF CLASSIFICATION FRAMEWORK

Labeled in-phase and quadrature (IQ) datasets captured using software defined radios (SDR). 12 classes comprised of 10 different transceivers plus 'noise' class and class for zero hz artifact from SDR direct conversion receiver.

The IQ data has the extension .sigmf-data, and the associated JSON file that contains the labeled metadata is .sigmf-meta.

There is a directory called 'data', this has all the extracted features for each class and is used for training.

There are also 12 folders named 'data_val1' - 'data_val12'. These contain the extracted features used validation accuracy.

There are numerous training, validation and testing datasets used to build the training and validation sets, all comprised of the ten primary RF modulation waveforms:

1) LoRa 125 (lora125)

2) LoRa 250 (lora250)

3) PTT analog transceivers (uv5r, vx8)

4) PTT digital transceivers (gd55, tyt)

5) RF doorbell (vod, sado)

6) misc. RF keyfob (click)

7) RF light switch (light)

8) Yaesu system fusion PTT (ysf)

The testing directory: testing_data_1msps conatins all the test data for the auto grader notebooks at the github repo.

These datasets go with the public github repo called: https://github.com/rdbadger/open_source_RFML