From Radar Cardiography to Electrocardiograms: Conditional Diffusion Model Enabled Contactless ECG Monitoring Using mmWave Radar

Cardiovascular disease (CVD) is one of the foremost causes of mortality globally, and cardiac arrhythmias constitute a major contributing factor. Continuous monitoring of cardiac signals plays a vital role for early detection and prevention. However, traditional electrocardiogram (ECG) devices require skin contact, which can be uncomfortable and inconvenient for long-term usage. In contrast, contactless cardiac health monitoring technologies, such as Wi-Fi and millimeter-wave (mmWave) radar, present a promising alternative. Millimete-rwave radar provides high range resolution, strong immunity to ambient light, and high sensitivity to small vibrations, making it ideal for contactless monitoring. However, mmWave radar primarily captures cardiac mechanical vibrations, known as radar cardiography (RCG) signals, which differ from the electrical activity recorded by clinical ECGs. To bridge this gap, we propose a contactless framework that uses mmWave radar and a conditional diffusion model to reconstruct ECG signals and then utilizes a deep learning model to classify arrhythmias. Specifically, mmWave radar captures RCG signals associated with cardiac activities. Leveraging the nonlinear relationship between cardiac mechanics and electrical activities, we design a Residual Network (ResNet)-based conditional diffusion model to convert these RCG signals into ECG signals. Finally, we develop a CNN-BiLSTM-SE network for arrhythmia classification. Experimental findings demonstrate the efficacy of the proposed approach for signal conversion as well as arrhythmia classification, offering a promising pathway toward contactless cardiac health monitoring.