PyRadarWF: A New Dataset Generator for Machine Learning of Radar Waveforms

Author(s):

Brent Kraczek, Matthew Ziemann, and Marius Necsoiu

Abstract:

Python Radar Waveform (PyRadarWF) is a new dataset generator for machine learning (ML) and analysis of radar waveforms. While ML based on deep learning has developed rapidly for 2-D image recognition, it lags behind for 1-D signal analysis, such as RF data and time-series data. This is true both of ML and the data needed to generate the ML models. Existing radar datasets are primarily proprietary, while open-source and DoD-available datasets are very limited. PyRadarWF has been developed to provide a wide variety of data for ML use, allowing customization to broaden the datasets. The data generation tools are designed to be extensible, including base waveform objects to allow for the straightforward addition of new radar waveforms, and a flexible data labeling format for training on the same data for different properties. In addition to data generation, PyRadarWF includes PyTorch-based ML modules and data plotting and analysis tools. We conclude with ML models trained on our generated data and tested against both emulated data and synthetic data from external sources.