Radar-Based Detection of Heart Sounds for Non-Contact Cardiac Timing: A Multi-Subject Evaluation Under Physiological Variability

Accurate estimation of cardiac mechanical events is essential for non-contact physiological monitoring in Internet of Things (IoT)-enabled healthcare systems and telehealth applications. In this study, we propose a radar-based approach for detecting the first heart sound (S1) and estimating its timing relative to electrocardiogram (ECG) signals within a remote sensing framework. The method integrates ECG-guided segmentation, envelope-based peak detection, and adaptive windowing to robustly identify S1 and S2 components from both radar-derived heart sounds and reference phonocardiogram (PCG) signals. The approach was evaluated on a publicly available dataset of synchronized radar and PCG recordings collected from eleven subjects under multiple physiological conditions. Under baseline conditions, the proposed method achieved a mean absolute error (MAE) of 25 ± 7 ms in estimating S1 timing relative to the ECG reference. Performance remained consistent across subjects, with errors ranging from approximately 13 ms to 32 ms, despite variations in signal quality and sensor placement. To assess robustness in realistic scenarios relevant to remote monitoring, additional evaluations were conducted under apnea_after_sport conditions. The method maintained comparable accuracy, achieving an MAE of 25 ± 10 ms, although increased variability was observed. The impact of physiological changes was found to be subject-dependent, with some subjects exhibiting improved performance while others showed degradation. These results demonstrate that radar-based heart sound detection can achieve millisecond-level accuracy and remains robust under physiological variability. The proposed approach highlights the potential of radar sensing as an IoT-enabled, non-contact solution for continuous cardiac monitoring, supporting scalable deployment in telehealth and remote patient monitoring applications.