Abstract
This dataset supplements the paper "Approaching Domain Generalisation with Embeddings for Robust Discrimination and Recognition of RF Communication Signals" and is composed as follows:
(1) The file "synthetic.zip" contains a total of 4,000 synthetically generated signals, created from 1,000 distinct synthetic wireless protocols. For each protocol, 4 signal examples were generated, each consisting of 16,384 samples. These signals represent the training data used for the RF signal embedding models.
(2) The file "vuorenmaa.zip" contains signal bursts of various types extracted from the dataset "Radio-Frequency Control and Video Signal Recordings of Drones" (https://doi.org/10.5281/zenodo.4264467; Authors: Miika Vuorenmaa, Jaakko Marin, Mikko Heino, Matias Turunen, & Taneli Riihonen; License: CC BY 4.0 - https://creativecommons.org/licenses/by/4.0/) and is part 1/3 of the evaluation data.
(3) The file "basak.zip" contains signal bursts of various types extracted from the dataset "Drone RF Dataset" (https://doi.org/10.48804/HZRVNZ; Authors: Sanjoy Basak, Sofie Pollin & Bart Scheers; License: CC BY-NC 4.0 - https://creativecommons.org/licenses/by-nc/4.0/) and is part 2/3 of the evaluation data.
(4) The files "files_train.json" and "files_eval.json" contain the splitting of the evaluation dataset used in section "4.2. Downstream classification task" of the paper.
(5) The file "plot_spectrograms.py" shows how to access the signal data using Python and visualizes the training and evaluation data. For the script to work, the contents of all *.zip files must be unzipped and located in the same folder as the Python script and the *.json files.
NOTE: Part 3/3 of the evaluation data has been moved to a separate dataset (https://fordatis.fraunhofer.de/handle/fordatis/461) due to incompatible licences.
Technical Information
SigMF (https://sigmf.org) is used as format for storing signal and metadata.