Millimeter-wave radar for non-contact heart rate monitoring: Composite filtering and multidimensional evaluation

Background Traditional heart rate monitoring relies on contact sensors, which limits its use in populations unsuitable for prolonged sensor attachment. Millimeter-wave radar (mmWave Radar) enables non-contact monitoring of vital signs, but its echoes are susceptible to respiratory and noise interference. Objective This study investigated the effects of different filtering methods on heart rate estimation and used electrocardiogram (ECG) as the reference to determine the optimal filtering configuration. Methods Five composite filtering methods were evaluated using the IWR6843AOPEVM 60 GHz radar platform. The Tri-Domain Similarity Index (TDSI), integrating time-domain, frequency-domain, and nonlinear features, was introduced to assess consistency between radar and ECG signals. Performance was evaluated through statistical analysis and robustness testing, and mmWave average heart rate was estimated using harmonic-based spectral scoring. Results The progressive composite filtering method achieved the best performance among the five methods, with a TDSI of 0.424. Robustness testing demonstrated stable performance under multiple noise conditions. Using harmonic-based spectral scoring for heart rate estimation, subject-level 5-fold validation yielded an overall mean absolute error (MAE) of 4.97 bpm. Conclusion This study presents a low-cost non-contact heart rate monitoring framework. The proposed progressive composite filtering method consisting of band-pass filtering (BPF), wavelet denoising (WD), and recursive least squares (RLS), together with multidimensional evaluation using TDSI, provided preliminary support for the feasibility of non-contact heart rate monitoring under controlled conditions.