Deep Learning for Radar Signal Detection in the 3.5 GHz CBRS Band
Raied Caromi, Alex Lackpour, Kassem Kallas, Thao T. Nguyen
This paper presents a comprehensive framework for generating radio frequency (RF) datasets, designing deep learning (DL) detectors, and evaluating their detection performance using both simulated and experimental test data. The proposed tools and techniques are developed in the context of dynamic spectrum use for the 3.5 GHz Citizens Broadband Radio Service (CBRS), but they can be utilized and expanded for standardization of machine learned spectrum awareness technologies and methods. In the CBRS band, environmental sensing capability (ESC) sensors are required to detect the presence of federal incumbent signals and trigger protection mechanisms when necessary. To support the development and evaluation of detection techniques for ESC sensors, we provide software tools for generation and augmentation of simulated radar datasets as well as baseline DL detectors that can be replicated, evaluated, and tested in a simulated or an experimental environment. We find that all the proposed detectors exceed ESC requirements for incumbent detection. The software tools, the pre-trained DL models and their configurations, and the experimental setup are made available in the public domain.
IEEE International Symposium on Dynamic Spectrum Access Networks
, Lackpour, A.
, Kallas, K.
and Nguyen, T.
Deep Learning for Radar Signal Detection in the 3.5 GHz CBRS Band, IEEE International Symposium on Dynamic Spectrum Access Networks, Virtual Conference, MD, US, [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=931843
(Accessed December 5, 2023)