Enhanced heart rate estimation method based on adaptive heart rate segmentation using millimeter-wave radar

In this paper, we propose an enhanced heart rate estimation method to solve the problem of large heart rate variations among individuals in home-based monitoring. For radar-based heart rate estimation, the primary challenges lie in the variability of heartbeat signals across individuals and their susceptibility to interference, such as respiratory harmonics. The proposed method addresses these challenges through two key strategies: (1) selecting heartbeat signals that exploit spatial diversity; and (2) precisely estimating heart rate using an adaptive heart rate segmentation network. Specifically, we first leverage the spatial diversity of heartbeat signals to extract potential heartbeat components in the localized chest region. Then, we divide heart rate range into many segments and employ a lightweight segmentation network to learn features of all segments. Finally, we refine the heart rate estimation using adaptive bandpass filters (BPFs) by the obtained results of the heart rate segmentation. The proposed method achieves a mean absolute error (MAE) of 1.06 beats per minute (BPM) and a root mean square error (RMSE) of 1.54 BPM. In addition, 93.80% of the real-time estimates within ± 2 BPM, and an intraclass correlation coefficient (ICC) of 0.8240.