IQ-Net: A Lightweight End-to-End Framework for Multi-Target People Counting Using Mmwave I/Q Signals
YaNan He, Yi Huang, Wei Hu, Qingfeng Zhang, Yaoqin Xie · 2025
Traditional radar counting methods perform well when targets are spatially separated but struggle in close-range, multi-person, stationary scenarios due to signal interference and overlap. We propose IQ-Net, a lightweight end-to-end neural network that directly processes raw I/Q data from millimeterwave radar, avoiding the need for handcrafted feature extraction. Experiments on 0–3 person counting tasks demonstrate that IQ-Net achieves an average accuracy of 93.95%, validating the effectiveness of raw I/Q signals in cluttered environments. The proposed approach offers a simple and efficient solution for closerange, low-cost multi-target radar sensing.