Presentation
13 June 2022 Time-frequency transforms for radar-based UAV recognition
Author Affiliations +
Abstract
Drone recognition has become a topic of increasing concern for defense applications. Due to the high speed of rotation of the drone blades, however, accurate drone recognition relies on sufficient time-frequency resolution of the drone radar micro-Doppler signature. Although one of the more commonly used time-frequency transforms is the spectrogram, such classical estimators embody a sub-optimal trade-off in temporal resolution versus frequency resolution. In this work, we evaluate the efficacy of various time-frequency transformations based on the latent space of deep neural networks (DNNs). In particular, we consider alternatives to the short-time Fourier transform, such as the wavelet transform, Wigner-Ville distribution, Choi-Williams distribution, and and super-resolution techniques, which have been recently shown to be effective on non-radar datasets, such as superlets. Transforms are compared for various millimeter wave radar systems for DNN-based classification.
Conference Presentation
© (2022) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Sean Kearney and Sevgi Zubeyde Gurbuz "Time-frequency transforms for radar-based UAV recognition", Proc. SPIE PC12108, Radar Sensor Technology XXVI, (13 June 2022); https://doi.org/10.1117/12.2619084
Advertisement
Advertisement
KEYWORDS
Time-frequency analysis

Transform theory

Temporal resolution

Unmanned aerial vehicles

Fourier transforms

Radar

Extremely high frequency

RELATED CONTENT


Back to Top