Abstract:Radar-based contactless cardiac monitoring has become a popular research direction recently, but the fine-grained electrocardiogram (ECG) signal is still hard to reconstruct from millimeter-wave radar signal. The key obstacle is to decouple the cardiac activities in the electrical domain (i.e., ECG) from that in the mechanical domain (i.e., heartbeat), and most existing research only uses pure data-driven methods to map such domain transformation as a black box. Therefore, this work first proposes a signal model for domain transformation, and then a novel deep learning framework called radarODE is designed to fuse the temporal and morphological features extracted from radar signals and generate ECG. In addition, ordinary differential equations are embedded in radarODE as a decoder to provide morphological prior, helping the convergence of the model training and improving the robustness under body movements. After being validated on the dataset, the proposed radarODE achieves better performance compared with the benchmark in terms of missed detection rate, root mean square error, Pearson correlation coefficient with the improvement of 9%, 16% and 19%, respectively. The validation results imply that radarODE is capable of recovering ECG signals from radar signals with high fidelity and can be potentially implemented in real-life scenarios.

What problem does this paper attempt to address?

### What problems does this paper attempt to solve? This paper aims to solve the difficult problem of non - contact reconstruction of electrocardiogram (ECG) signals from millimeter - wave radar signals. Specifically, the main challenge of the research lies in how to decouple the activities of the heart in the electrical domain (i.e., ECG) from those in the mechanical domain (i.e., heartbeat) and generate high - fidelity ECG signals. #### Main problems: 1. **Decoupling heart activities**: - There are complex interactions between heart activities in the electrical domain and the mechanical domain. However, most of the existing methods use a pure data - driven approach for domain conversion, which makes the decoupling process like a black - box operation, difficult to explain and optimize. 2. **Improving reconstruction accuracy and robustness**: - Existing methods perform poorly when dealing with sudden noises such as body movements. These noises are usually several orders of magnitude larger than heart - related vibrations, which disrupts the forward propagation of deep neural networks and affects the reconstruction quality of ECG signals. 3. **Lack of effective signal models**: - At present, there are no easy - to - understand signal models that can be integrated into domain conversion to guide the conversion from radar signals to ECG signals. #### Solutions: To solve the above problems, this paper proposes a new deep - learning framework - **radarODE**, with the following main features: - **Introducing ordinary differential equations (ODE)**: By embedding ODE as a decoder, it provides morphological prior knowledge, helps the model training converge, and improves the robustness to noises such as body movements. - **Designing a single - cycle ECG generator (SCEG)**: Based on the proposed signal model, the SCEG module parameterizes radar signals into sparse representations with morphological significance, enhances the characteristic peaks of the recovered ECG signals, and accelerates the model training process. - **Fusing morphological and temporal features**: By fusing the morphological and temporal features extracted from the original radar signals, the radarODE framework can achieve more accurate ECG signal reconstruction. #### Verification results: Through verification on public datasets, radarODE has improved by 9%, 16% and 19% respectively in terms of the miss - detection rate, root - mean - square error (RMSE) and Pearson correlation coefficient, which proves its superior performance in high - fidelity ECG signal reconstruction. In summary, this paper successfully solves the key problem of non - contact reconstruction of high - fidelity ECG signals from millimeter - wave radar signals by introducing ODE and designing a new deep - learning framework, and shows its potential in practical application scenarios.

radarODE: An ODE-Embedded Deep Learning Model for Contactless ECG Reconstruction from Millimeter-Wave Radar

radarODE-MTL: A Multi-Task Learning Framework with Eccentric Gradient Alignment for Robust Radar-Based ECG Reconstruction

RadarNet: Noncontact ECG Signal Measurement Based on FMCW Radar

Contactless Electrocardiogram Monitoring with Millimeter Wave Radar

Human Recognition with the Optoelectronic Reservoir-Computing-based Micro-Doppler Radar Signal Processing.

ECG waveform generation from radar signals: A deep learning perspective

Cardiogram Detection with a Millimeter-wave Radar Sensor

HeRe: Heartbeat Signal Reconstruction for Low-Power Millimeter-Wave Radar Based on Deep Learning

Contactless Cardiogram Reconstruction Based on the Wavelet Transform Via Continuous-Wave Radar

RF-Heartbeat: Robust and Contactless Heartbeat Monitoring Based on FMCW Radar

Radar-Beat: Contactless beat-by-beat heart rate monitoring for life scenes

Deep Learning Radar for High Fidelity Heart Sound Recovery in Real-World Scenarios

An Overview of Algorithms for Contactless Cardiac Feature Extraction From Radar Signals: Advances and Challenges

Detection Of Bio-Signals From Body Movement Based On High-Dynamic-Range Doppler Radar Sensor (Invited)

Contactless Cardiac RR Intervals Estimation in Radar-Based Cardiogram Detection

Radar-Based High-Accuracy Cardiac Activity Sensing

AirECG: Contactless Electrocardiogram for Cardiac Disease Monitoring via mmWave Sensing and Cross-domain Diffusion Model

Classification of Hemodynamics Scenarios from a Public Radar Dataset Using a Deep Learning Approach

Contactless Radar Heart Rate Variability Monitoring Via Deep Spatio-Temporal Modeling