<p>Heart rate variability (HRV) plays a critical role in the prediction of diseases related to cardiovascular and autonomic nervous system. HRV monitoring using millimeter-wave radar has been exploited as a promising research direction for daily health monitoring and supervision of patients with skin diseases. A critical issue with these solutions includes that good monitoring performance could only be achieved with unmixed heartbeat signal, while random body movements (RBM) and interference from respiration harmonics severely hinder the separation of heartbeat signal. In this paper, we propose a millimeter-wave radar-based HRV monitoring system to mitigate interference and significantly improve HRV detection accuracy. With careful noise filtering, the system is able to minimize the interference caused by RBM. To address the respiration harmonics issue, we design a heartbeat signal separator based on a variational problem model to extract the precise heartbeat signal. Comprehensive experiments demonstrate that the system can achieve a median Interbeat Intervals (IBI) estimation error of 24 ms. For the challenging multi-user coexistence scenario, the system is able to achieve the median IBI estimation error of 28 ms and 32 ms for the two-user and three-user coexistence scenarios.</p>