Aims. Remote heart rate monitoring using Doppler radar is challenged by heart rate variability (HRV), respiratory motion and low Signal-to-Noise Ratio (SNR) dominated by close-in phase noise. This work investigated the enhancement of cardiac component extraction through a non-uniform temporal alignment algorithm designed to compensate for HRV.
Methods. Measurements were conducted on seven adults (N=7) employing a 2.4 GHz Continuous Wave Doppler radar alongside a Pulse Oximeter (PO) as reference. After band-pass filtering to isolate the cardiac band, the algorithm extracted the instantaneous cardiac phase of each recording via the Hilbert transform, resampling the Doppler data onto a phase-uniform grid to enforce the periodicity of heartbeat cycles. Two recalibration configurations were assessed: a PO-aided mode and a fully autonomous blind mode. Each was then evaluated under both voluntary apnoea and free-breathing conditions.
Results. Under apnoea, alignment focused cardiac energy into a narrowed frequency band, yielding SNR gains up to 5.7 dB and revealing higher-order harmonics previously obscured by noise. The blind mode achieved a heart rate root mean square error of 5.1 beats per minute relative to the PO reference. During free-breathing, PO-aided alignment achieved a 7.1 dB mean gain, peaking over 11 dB in two of the seven subjects and recovering cardiac harmonics undetectable in raw signals. Conversely, the performance of the blind approach degraded due to unreliable cardiac phase estimation, with a maximum gain of only 3.8 dB observed on a single subject.
Conclusion. Non-uniform temporal alignment effectively restores the cardiac harmonic structure in Doppler radar sensing. While robust during apnoea and PO-aided cases, autonomous extraction under free-breathing remains an open challenge. These findings identify parametric respiratory modelling as the key remaining challenge for fully autonomous blind separation.