Contactless Radar Applications
Contactless Radar Applications
Non-contact vital signs monitoring (NCVSM) of multiple people in noisy, cluttered environments
Background:
Non-contact vital signs monitoring (NCVSM) of multiple individuals, such as respiration and heartbeat, has been investigated in recent years due to the rising cardiopulmonary morbidity, the risk of disease transmission, patient discomfort and heavy burden on medical staff. Frequency Modulated Continuous Wave (FMCW) radars have shown great promise in meeting these needs. However, contemporary techniques for NCVSM via FMCW radars are based on simplistic models and present difficulties in coping with noisy environments containing multiple objects.
Sparsity-Based Multi-Person Localization and NCVSM
We developed an extended model for multi-person NCVSM via FMCW radar. By utilizing the sparse nature of the modelled signals in conjunction with human-typical cardiopulmonary features, we present accurate localization and NCVSM of multiple individuals in a cluttered scenario, using only a single channel. Specifically, we provide a joint-sparse recovery (JSR) mechanism to localize humans in a noisy environment and develop a robust method for NCVSM called VSDR (for vital signs based dictionary recovery), which uses a dictionary-based approach to search for the rates of respiration (RR) and heartbeat (HR) over high-resolution grids corresponding to human cardiopulmonary activity.
Fig. 1. Block diagram of the proposed approach for multi-person NCVSM via SISO FMCW radar, which yields accurate estimates of RR and HR by exploiting the sparse composition of the input data in accordance with human cardiopulmonary properties, using only a single channel.
Simulated analysis:
The simulated trials combine the proposed model with in-vivo data of 30 monitored individuals. We demonstrate accurate human localization in a noisy scenario that includes both static and vibrating objects and show that our VSDR approach outperforms existing NCVSM techniques, based on several statistical metrics.
A. Multi-Person Localization in Cluttered Scenarios
cenario #1:
Multiple objects located at different radial distances – using a SISO setup
Fig.2. Multi-person localization by the setup of Table I, for SNR = 0[dB]. (a) The Range vs. Slow-Time map. The row-wise intensities correspond to the DC component and the reflections of Table I 's objects. (b) Several localization techniques based on the average power [4], the std estimate.
[5] and the proposed JSR [1]. Only JSR localization properly detects the subjects in the given scenario.
Scenario #2:
Multiple objects located at the same radial distance but at different angles – using a SIMO setup
Fig. 3. Multi-person localization by the setup of Table 1 that includes equidistant targets, for SNR=0 [dB]. Only the proposed JSR method properly detects the humans in the given scenario.
B. Multi-Person NCVSM.
Fig. 4. Simultaneous NCVSM of 3 subjects given their thoracic vibrations, w.r.t. known references, SNR=0 [dB].
Rows: 3 different subjects. Columns: Extracted thoracic vibrations for a given time interval, proposed VSDR [1], FFT w/ ZP, FFT w/o ZP and Phase-Reg. The VSDR estimates are the closest to the reference estimates for both RR and HR
In terms of performance, the VSDR outperforms the other compared methods in 4 different metrics, for any examined SNR case.Specifically, In the more challenging task of HR monitoring, the gap between the VSDR and the other techniques is prominent.
Fig. 5. Median NCVSM performance of the proposed VSDR [1] compared to other techniques vs. SNR. Rows: HR, RR. Columns: SR - 2 bpm, PCC, MAE, RMSE.
Experimental analysis:
https://www.youtube.com/watch?v=7BCgcNp_BA4
Hardware
We developed a dedicated hardware phantom that simulates multi-person thoracic displacements through which realistic experiments for NCVSM using radars can be performed. The phantom reduces the need for human trials and enables proper calibration of the suggested algorithms.
Our demonstration platform consists of (1) a vibration generator for imitating thoracic displacements, (2) flat circular metal plates (24 cm diameter), (3) a radar sensor, located up to 2 meters from the phantom, and (4) a dedicated experimental setup.
Scenario #1:
Hardware phantom for single-person NCVSM via radar
Graphical User Interface:
Vhttps://www.youtube.com/watch?v=JMqkVA4_jQE
Scenario #2:
Hardware phantom for multi-person NCVSM via radar.
Graphical User Interface
Example of localization and NCVSM of 3 people simultaneously:
Two equidistant targets located approximately at 1 [m] from the radar and +-10 [deg]. The third target is located approximately at 1.1 [m] from the radar and +25 [deg] .
The operator can choose between a trial of sinusoids or human thoracic vibrations as well as determining the frame rate and duration of the trial
Videos:
ICASSP22 Hardware Demo: Contactless Vital Signs - YouTube.
https://www.youtube.com/watch?v=VzhkbVaiR4M
Sparsity-Based Multi-Person NCVSM Using Radar - YouTube.
https://www.youtube.com/watch?v=F7O5gQG0D94
ICASSP 2023 IEEE International Conference on Acoustics, Speech and Signal Processing:
Articles from the media
https://www.ynet.co.il/environment-science/article/bytggrhkgg#google_vignette
https://www.israel21c.org/noncontact-radar-unit-monitors-patients-vitals-remotely/
https://www.jpost.com/health-and-wellness/article-853525
Extension to NCVSM during large movements (such as walking)
- Real-world healthcare scenarios, e.g., patients shifting in bed, or monitoring uncooperative subjects, require robust non-contact vital sign monitoring (NCVSM) during movement
- Proposed solution for moving people using SISO FMCW radar: extends [1] by
- Sparsity-based localization scheme involving Kalman filtering for tracking people during large movements
- Signal separation framework leveraging cardiopulmonary patterns and movement dynamics
- The Extended Vital Signs-based Dictionary Recovery (E-VSDR) [5] method is then used to continuously estimate the heart rate (HR)
- Our signal separation approach yields superior NCVSM results compared to state-of-the-art separation methods.
Video:
https://www.youtube.com/watch?v=ASzNfUx6k7Y
RaLU-Net: Deep unfolded radar localization of humans
- Rising demand for monitoring solutions in crowded environments
- Detecting and positioning multiple humans with radar is highly affected by resolution limitation
- This work first provides the iterative localization method: RaLU-JSR
Then, it is unfolded into a neural network: RaLU-Net
- Utilizes the unique structure of the data to enhance accuracy and reduce computational cost
- Achieving high performance with minimal computation in thousands of close-proximity scenarios
Non-contact Spirometry test
SpiRadar is a non-contact spirometry framework that utilizes mmWave FMCW radar to deliver comprehensive pulmonary assessments without the discomfort or compliance barriers of traditional mouthpieces. By integrating physiologically-motivated signal preprocessing with a feature-dependent polynomial model, the system translates millimeter-scale thoracic displacements into complete spirometric volume and flow curves. Leveraging sparse optimization, the framework is designed to generalize directly to new subjects without the need for individual device calibration. Validated on a diverse pediatric cohort, the system reliably estimates essential clinical parameters and assesses bronchodilator treatment responses. This approach offers a practical, patient-friendly alternative for respiratory monitoring that is well-suited for pediatric clinics, remote patient care, and environments requiring minimal physical contact.
Non-Invasive Quantitative Monitoring of Nanoparticles In Vivo Using Millimeter Wave Radar
The identification of nanoparticles in specific regions of the body can indicate the presence of pathological conditions such as tumors, metastases, or infections. This occurs either due to natural tendency of nanoparticles to accumulate in areas with enhanced permeability and retention (EPR effect), or through active targeting mechanisms involving conjugation to specific
monitor changes in their concentration over time—has significant potential as a non-invasive diagnostic and monitoring approach.
We further proposed HM-ISTANet, a physics-informed deep unfolding-based approach for H-field electromagnetic direct inversion, designed for superparamagnetic nanoparticle (SPION) imaging. To resolve background field distortions caused by heterogeneous tissue impedance and dynamic magnetic modulations, we incorporate a two-stage adaptive H-field correction mechanism. Furthermore, a channel-wise Group Normalization scheme is introduced within the unrolled cascade to prevent global batch-statistic drift and ensure training stability under physics-informed constraints. Both numerical simulations and experimental evaluations demonstrate that our method outperforms classical electromagnetic inversion baselines (such as Born approximation and Full Waveform Inversion) in reconstruction quality and structural resolution under noisy conditions, while maintaining a high level of computational efficiency
(ICASSP), 2024, pp. 2066-2070.
Y. Eder, Y. Kvich, R. Guo, and Y. C. Eldar, "RaLU-Net: Deep Unfolded Radar Localization of Humans for Precise Multi-Person Non-Contact Vital Signs Monitoring," in Proc. IEEE Int. Conf.
Contact Free Blood Pressure Waveform Estimation from Radar Signals via Multimodal Dictionary Learning
❯ The Need for Contactless BP: Continuous BP monitoring is vital for cardiovascular health, but gold standard A line is invasive, while traditional cuffs cause discomfort and limit long-term use.
❯ What we do: We investigate contact-free blood pressure (BP) monitoring using a chest radar. We propose a dictionary-learning framework that leverages multimodal recordings including radar, ICG, ECG, and BP to learn shared physiological representations. After training, BP waveforms are reconstructed by weighting the learned BP dictionary with sparse codes estimated solely from the radar signal. The learned representation enables interpretable cardiac coding from radar signals alone, allowing the synthesis of high-fidelity BP waveforms.
Dictionary learning results
References:
[1] Y. Eder and Y. C. Eldar, "Sparsity-Based Multi-Person Non-Contact Vital Signs Monitoring via FMCW Radar," IEEE Journal of Biomedical and Health Informatics, vol. 27, no. 6, pp. 2806-2817, 2023.
[2] Y. Eder, Z. Liu, and Y. C. Eldar, "Sparse Non-Contact Multiple People Localization and Vital Signs Monitoring Via FMCW Radar," in Proc. IEEE Int. Conf. Acoust., Speech, Signal Process. (ICASSP), Rhodes Island, Greece, 2023, pp. 1-5.
[3] Y. Eder, R. Abel, A. Schroeder, and Y. C. Eldar, "Localization and Tracking of Gold Nanoparticles Using mmWave FMCW Radar," in Proc. IEEE Int. Conf. Acoust., Speech, Signal Process.
coust., Speech, Signal Process. (ICASSP), April 2025, pp. 1-5.
[4] Y. Eder, E. Zagoury, S. Savariego, M. Namer, O. Cohen, and Y. C. Eldar, "Robust Phantom-Assisted Framework for Multi-Person Localization and Vital Signs Monitoring Using MIMO FMCW Radar," arXiv preprint, arXiv:2501.06755, 2025.
[5] Y. Eder, Y. Kvich, R. Guo, and Y. C. Eldar, "RaLU-Net: Deep Unfolded Radar Localization of Humans for Precise Multi-Person Non-Contact Vital Signs Monitoring," in Proc. IEEE Int. Conf. (ICASSP), 2024, pp. 2066-2070.
[6] M. Xu, H. Ruan, R. Guo and Y. C. Eldar, "Contact Free Blood Pressure Waveform Estimation from Radar Signals via Multimodal Dictionary Learning", IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP), May 2026.