Radar systems improve heart signal recovery with fewer resources
Kapture: Capturing Cardiac Dynamics with Koopman-Governed Learning for Efficient Radar-Based Electrocardiogram Recovery
Artificial Intelligence
Summary
Measuring heart activity without touching a person is tricky because other movements and noises mix with the heart signals. The paper presents Kapture, a new method that uses math to simplify and clean up radar data to better capture heartbeats. This approach uses less computing power and fewer settings while still accurately showing heart signals. It works well even when compressed, making it more efficient for devices to monitor heart health quietly.
What this means in practice
- •For medical device engineers: Build more compact and power-efficient radar systems for remote heart monitoring that maintain signal accuracy despite noise and interference.$Commercial implications: Enables medical devices that use radar for contactless ECG monitoring with fewer hardware and computational resources, suitable for consumer or clinical products.
- •For wearable health tech developers: Incorporate efficient algorithms to recover heart signals from radar data, improving wearable devices’ ability to track cardiac health without sensors attached to the skin.
Authors
Tong Wu, Jing Peng, Ziqi Feng, Yuanyuan Zhang
Abstract
Millimeter-wave (mmWave) radar enables unobtrusive, contactless electrocardiogram (ECG) reconstruction for cardiac monitoring. Time-frequency spectrograms preserve fine cardiac patterns but often require large backbones to separate ECG-relevant features from respiration, motion, multipath, and subject-dependent interference. We propose Kapture, a parameter-efficient Koopman-governed framework that projects radar hidden states into a low-dimensional observable space and identifies a regularized full linear evolution operator from adjacent observable states. The Koopman-predicted observables refine subsequent hidden states for ECG reconstruction. To suppress predictable interference dynamics, a temporal contrastive objective pulls neighboring states together and separates non-neighbors, while reconstruction supervision preserves task relevance. Using approximately 80 minutes of quasi-static radar-ECG recordings containing realistic noise from body movements and other sources, Kapture consistently improves reconstruction across matched backbone widths, with the largest gains under aggressive compression. The compact configuration approaches the full-width reference accuracy with 68.1% fewer parameters and 91.3% fewer profiler-covered floating-point operations (FLOPs), while the full-width configuration delivers the strongest overall reconstruction performance. Our code will be made publicly available after potential publication.