GNSS Interference Spectrum Highway Dataset 1

Introduction

We publish the Spectrum Highway Dataset 1 that was recorded on a bridge over a German highway and contains a variety of interference classes in GNSS signals from spectral recording hardware. For the dataset recording, we developed a hardware setup that captures short, wideband snapshots in both E1 and E6 GNSS bands. This setup is mounted to a bridge over a highway. The setup records 20ms raw IQ snapshots triggered from the energy with a sample rate of 62.5MHz, an analog bandwidth of 50MHz and an 8bit bit-width.

The dataset can be used for research purposes to addresse the following research problems:

• GNSS interference detection
• GNSS interference classification with 8 interference classes
• Adaptation to new interference classes by addresses the following ML topics:
  
  • Domain adaptation
  • Few-shot learning
  • Continual learning
• Interference mitigation
• ML architecture search
• Addressing robustness of ML models in unbalanced datasets

Dataset

The folder GNSS_dataset contains the Spectrum Highway Dataset 1 with the corresponding labels in .txt files for the train-(adaptation-)test split. To use the dataset extract all .npy files from the .zip files. One sample in the text file is defined as for example:

data/157369.npy 0

For classical ML methods, use the train.txt and test.txt dataset to train and evaluate your model. For methods such as few-shot learning, use the train.txt dataset for the baseline training, the FSL_support_set.txt as the support datset, the FSL_adaptation_set.txt as the adaptation dataset, and FSL_adaptation_test.set for the final testing dataset. The test.set is split into the FSL_adaptation_set.txt and FSL_adaptation_test.set datasets.

At certain frequencies the GPS/Galileo or GLONASS signals can easily be seen as a slight increase in the spectrum. Note that experts manually analyzed the datastreams by thresholding CN/0 and AGC values. Manual labeling of these snapshots has resulted in 11 classes: Classes 0 to 2 represent samples with no interferences, distinguished by variations in background intensity, while classes 3 to 10 contain different interferences.

License

This work is licensed under a CC BY-NC-SA 4.0: Creative Commons Attribution-Noncommercial-ShareAlike, see https://creativecommons.org/licenses/by-nc-sa/4.0/

Further Datasets & Methods

For more information on the dataset, links to other datasets, and our methodologies, visit the following Gitlab page: https://gitlab.cc-asp.fraunhofer.de/darcy_gnss