SkeletonHR: Robust Heart Rate Detection Under Movement Conditions Using a Commodity mmWave Radar Sensor

FMCW mmWave radar has shown promise in vital signs measurement in stationary scenarios. However, its precision for heart rate detection significantly degrades in moving scenarios due to overwhelming action noise. To address this issue, we introduce SkeletonHR, a novel approach for robust human heart rate detection in complex movement scenarios using a commodity mmWave radar. SkeletonHR first leverages an advanced network for human joint points reconstruction, and four additional joint points were interpolated to fully extract the phase signals from the upper body. For the purpose of isolating the complete heartbeat signal from the complex composite signal, we propose a HeartWave-variational mode decomposition (VMD) algorithm that extracts the heartbeat signal by rigorously evaluating its physiological validity and quality of the heartbeat signal. Furthermore, an adaptive joint-point prioritization strategy is explored to select the most reliable signal across multiple chest joints to enhance the accuracy and robustness of heart rate detection. Ablation experiments demonstrate the effectiveness of our proposed method. Experimental results show that SkeletonHR can accurately separate heartbeat in the presence of micro and large random body movements. The 80th percentile of the absolute heart rate error per minute is less than 4.51 BPM in the case of micro movement and 10.35 BPM in the case of large movement.