Low-Latency Spike-Based Range and Velocity Estimation of FMCW Radar Signals

Abstract

Frequency-Modulated Continuous-Wave (FMCW) radars determine a target’s range, velocity, and angle of arrival by performing multiple Fourier analyses on received signals. However, this processing is conventionally frame-based, requiring waiting for an entire frame of data to be stored in memory and processed. In this work, we propose an event-based approach to two-dimensional Fast Fourier Transform (FFT) radar processing using Spiking Neural Networks (SNNs). Unlike standard pipelines that demand large data buffers for range-Doppler analysis, our method operates chirp-by-chirp, thus allowing for low-latency estimates. Using mathematical derivations and computer simulations, we demonstrate the same performance of a traditional 2D FFT processing pipeline, while offering a viable event-based alternative to conventional frame-based solutions for FMCW radar systems.