MM-FGHM: Fine-Grained Heartbeat Monitoring Using MIMO Millimeter-Wave Radar.

In recent years, radar systems have gained attention in the field of health monitoring, driven by the availability of low-cost radar equipment and the development of efficient algorithms. Reconstructing continuous and complete fine-grained heartbeat waveforms is essential, as it provides detailed insights into cardiac activity that a single heart rate value cannot capture. And the final performance of heartbeat waveform reconstruction is determined by the synergy between signal preprocessing and architectural selection. In this paper, we propose MM-FGHM, a contactless radar-based heartbeat monitoring system, achieving fine-grained heartbeat waveform reconstruction from the radar signal and accurate cardiac metric estimation. Specifically, we design a dual-stream network termed ResED-Net, which integrates ResNet with an encoder-decoder architecture to fully extract heartbeat features from the real and imaginary parts of 3D Range-Angle-Time matrices. Additionally a joint loss function is also proposed to facilitate high-precision reconstruction. Comprehensive experiments with the data from 16 users across 9 spatial configurations in multiple scenarios validate the system's performance. Results demonstrate that MM-FGHM achieves high accuracy in heartbeat waveform reconstruction and cardiac metric estimation, also with good robustness and generalization, which indicates its potential for reliable non-contact health monitoring.