Abstract: In recent years, health issues have become serious worldwide. As a vital indicator for evaluating human health, heart rate (HR) detection, which comes without disturbance and with comfort, has become an urgent need of society. Traditional detection methods in medical institutions, such as photoplethysmography and electrocardiography, although effective in providing real-time and accurate data, suffer limitations in comfort and versatility. Advances in radar technologies enable noncontact detection of HR. However, as the chest wall displacement caused by the heartbeat is extremely weak, HR can be easily overwhelmed by respiration harmonics, noise, and clutter. Hence, two critical challenges arise during the detection process: unknown environmental noise and respiratory harmonic interference. To achieve accurate HR estimation without disturbance, we propose a noncontact HR detection approach using a millimeter-wave radar based on the Hilbert–Huang transform (HHT) and spectrum-weighted reconstruction. The approach includes a micromotion target localization strategy and an HR reconstruction estimation strategy. In the micromotion target localization strategy, we first eliminate static clutter in the raw data. Thereafter, building on the traditional constant false alarm rate (CFAR) method, we design an adaptive CFAR approach that dynamically adjusts based on environmental noise thresholds, incorporating real-time scaling factor updates to reduce the impact of random dynamic noise, thereby enhancing the sensitivity and accuracy of weak signal target detection during radar-based HR monitoring. In addition, due to the periodic nature of physiological signals in the thoracic region and the relatively random nature of interference, autocorrelation analysis is employed for periodic identification, further reconfirming the target positions and enhancing the accuracy of localization. In the HR signal reconstruction strategy, we first use HHT for high-resolution time-frequency localization analysis, capturing transient features and variations in nonstationary and nonlinear signals such as those of heartbeats. By extracting the intrinsic mode functions corresponding to the HR range and designing a spectral weighting reconstruction method, we segment and enhance the HR range, because of which respiratory harmonics and noise interference in the heartbeat signal are further suppressed, thereby improving the resolution of HR detection. Experiments are conducted in laboratory and office settings using the Texas Instruments (TI) IWR1843 millimeter-wave radar sensor, involving 10 different participants to evaluate the effects of HR, distance, angle, and user heterogeneity on HR detection. To assess the effect of varying distances on HR estimation, five distances are selected: 0.5, 1.0, 1.5, 2.0, and 2.5 m, with participants positioned facing the radar. The absolute error of the proposed HR estimation method increases from 0.007 to 0.026 Hz as distance is increased. The effect of different angles on HR estimation is also analyzed at 0°, 15°, and 30°, showing that the signal-to-noise ratio of radar echo signals decreases as distance and angle are increased, resulting in increased absolute error in HR estimation. Furthermore, a comparative analysis is performed between the proposed method and three commonly used methods: variational mode decomposition (VMD), ensemble empirical mode decomposition (EEMD), and zero-attracting sign exponentially forgetting least mean square (ZA-SEFLMS). The proposed method outperforms the other methods in HR estimation by effectively suppressing respiratory harmonics and environmental noise clutter, resulting in superior decomposition and reconstruction of the heartbeat signal, with an average HR error of 0.019 Hz, considerably enhancing HR detection accuracy.