Radar-based contactless human ECG signal reconstruction

With the increasing application of radar in non-contact vital signs monitoring, researchers have extensively explored methods to extract respiratory and heartbeat information from radar echoes to assess human health. However, current research primarily focuses on accurately estimating respiration and heart rates, which cannot meet the need of comprehensive health monitoring. Additionally, medical professionals do not directly analyze patient conditions through radar signals, but with electrocardiogram (ECG) signals. Thus, this paper aims to explore the mapping relationship between radar echoes and ECG signals to further reveal the vital sign information contained therein. A deep learning model based on complex-valued neural networks and self-attention mechanism is proposed to realize the conversion of radar signals and corresponding ECG signals. The experiment result shows that the proposed method can successfully convert radar vital signs into ECG signals, with an average Pearson correlation coefficient of 0.98.