CFARNet: Learning-Based High-Resolution Multi-Target Detection for Rainbow Beam Radar

TL;DR

This work addresses multi-target detection in mmWave rainbow-beam OFDM radar, where conventional CFAR-based peak detection limits resolution. It replaces CFAR with CFARNet, a CNN that predicts peak subcarriers in the angle-Doppler domain, and couples this with MUSIC for high-resolution range and radial velocity estimation. Across simulations, CFARNet substantially outperforms a CFAR+MUSIC baseline, especially for closely spaced targets and low transmit power, achieving superior angular resolution and robustness with faster inference. The results demonstrate the viability of data-driven peak detection for real-time ISAC radar sensing in dense, challenging environments.

Abstract

Millimeter-wave (mmWave) OFDM radar equipped with rainbow beamforming, enabled by phase-time arrays (PTAs), provides wide-angle coverage and is well-suited for fast real-time target detection and tracking. However, accurate detection of multiple closely spaced targets remains a key challenge for conventional signal processing pipelines, particularly those relying on constant false alarm rate (CFAR) detectors. This paper presents CFARNet, a learning-based processing framework that replaces CFAR with a convolutional neural network (CNN) for peak detection in the angle-Doppler domain. The network predicts target subcarrier indices, which guide angle estimation via a known frequency-angle mapping and enable high-resolution range and velocity estimation using the MUSIC algorithm. Extensive simulations demonstrate that CFARNet significantly outperforms a baseline combining CFAR and MUSIC, especially under low transmit power and dense multi-target conditions. The proposed method offers superior angular resolution, enhanced robustness in low-SNR scenarios, and improved computational efficiency, highlighting the potential of data-driven approaches for high-resolution mmWave radar sensing.

Figures (6)

• Figure 1: System architecture of phase-time array rainbow beamforming for multi-target sensing.
• Figure 2: An example of the angle-Doppler log-magnitude spectrum $\mathbf{L}_{\text{AD}}$. There are 3 targets, $M=2048$ subcarriers and $32$ Doppler indices.
• Figure 3: Architecture of the proposed CFARNet network.
• Figure 4: An example of CNN-based CFARNet output vs. CFAR output vs. Ground truth when there are three close-by targets.
• Figure 5: 90th percentile error performance vs. transmit power for CFARNet and YOLO baseline across different minimum angular separations $\Delta\phi_{\text{min}} \in \{1^\circ, 1.5^\circ, 3^\circ, 5^\circ, 10^\circ\}$.