VitalSense - Performance Validation of mmWave Radios Based Trusted Biometric Intelligence Sensing

Robust Biometric Information Sensing With mmWave Radar System-on-Chip

     

Ruochen Wu ^✉     Laura Miro ,     Albert Aguasca

Montse Najar     Antoni Broquetas ^✉

* The CommSensLab-UPC and the Signal Processing and Communications Group are recognized consolidated research groups (GRC-01415 and 2021 SGR 01033) by the Generalitat de Catalunya.

** This work has been supported by the Spanish Ministry of Science, Innovation and Universities MICIU/ AEI/10.13039/501100011033 and the European Regional Development Fund FEDER, UE, with projects PID2020-117303GB-C21, PID2022-138648OB-I00, and PID2024-161188OB-C21, the China Scholarship Council (CSC) under Grant 202208390068, and the Industrial Doctorates Plan of the Department of Research and Universities of the Generalitat de Catalunya.

English | 中文 | Español

---

Radio remote sensing and millimeter-wave (mmWave) sensing solution based on 120 GHz Frequency-Modulated Continuous Wave (FMCW) Radar System-on-Chip (RSoC) for smart healthcare monitoring, Internet of Medical Things (IoMT), and biometric extraction.

Note

We are currently collaborating with the Hospital Universitari Germans Trias i Pujol (HUGTiP) (Institut de Recerca Germans Trias i Pujol, IGTP) of Barcelona to carry out an experimental validation of the developed mmWave Radar for Vital Sensing on patients of the Cardiology Service.

Previous Collaborative Project (with Hospital Sant Joan de Déu Barcelona) Introduction Video

VitalSense2018.mp4

Important

The study has been approved by the ethics committee of the Universitat Politècnica de Catalunya · BarcelonaTech (Identification code: 2024-028). All subjects provided their informed consent to voluntarily participate in this study.

RSoC for Wireless Sensing

mmWave Radar Sensor Prototype

The used non-commercial RADAR has been conceived, designed and built in our laboratory (CommSensLab-UPC) specifically for the intended applications.

Parameter	Value
Center Frequency ($f_{0}$)	122.5 GHz
Radar Nominal Bandwidth ($B$)	1 GHz (in the ISM band)
Antenna Beamwidth ($\theta_{\text{3dB}}$)	$2^{\circ}$
Radar Range Resolution ($\Delta r$)	$\frac{c}{2B}=$ 150 mm
Wavelength ($\lambda$)	$\frac{c}{f_{0}}=$ 2.449 mm
Pulse Repetition Period ($T_{\text{frame}}$)	3 ms
Chirp Slope Time ($T$)	1.5 ms

Warning

The radar bandwidth can be programmed up to 4 GHz. In our experimental setup, a radar bandwidth of 3 GHz was configured.

Files for hardware...

Radar measurement: AlazarTech

ATS-SDK is a Windows and Linux compatible software development kit created by AlazarTech to allow users to programmatically control and acquire data from its line of waveform digitizers, which fully supports for C/C++ and C# (Visual Studio or GCC), MATLAB, LabVIEW and Python environments. In this case, we complete the project based on MATLAB.

Dataset

Vital Signals Database - acquired by CommSensLab (Dept. of Signal Theory and Communications) (Internal experimental data)

Tip

📣 A new radar vital signals DATASET comprising 24 healthy subjects has been published! 🎉

Sample data

Several sets of sample vital signal data in the data are used for testing, familiarization and studying of the algorithm.

Data file naming rules: "SUBJECT + MEASUREMENT POSITION + STATE + with ECG (optional) .mat"

Signal Processing Algorithm

MAIN FILE: main

Important

Be sure to check the parameter settings and read the relevant comments before running!

Achievements

Vital sensing radar with intelligent adaptive multi-phase signal processing chain to deliver for each monitored subject three complementary types of information:

• An adapted filter perfectly matched to the monitored subject radar cardiac pulse waveform, providing the best possible Signal to Noise Ratio and interference rejection.

• The repetitive radar blood pressure waveform estimation, which is not only an additional biologicalcharacteristic for biometrics, but also an alternative to conventional invasive/contact sensors in determining the condition of the cardiovascular system.

• The robust detection and precise temporal alignment of the cardiac pulses allowing to accurately measure heart-rate and to detect anomalies, resulting in more precise biometric parameters.

• The developed novel "white-box" technique replaces data-hungry AI black-box models to achieve the privacy-preserving wireless sensing. This breakthrough enables secure and robust biometric information identification at the edge, establishing a foundation for endogenous safety and security and next-generation confidential computing in IoMT.

---

Workflow

1. Signal Preprocessing

• Vital signal $s_{vital}$ obtaintion by phase unwrapping
• Signal separation: respiratory signal $s_{b}$ extraction with FIR linear-phase filter; cardiac signal -> $s_{h}=s_{vital}−s_{b}$

2. Real-time Repetitive Waveform Adaptive Matched Filter (RWAMF)

• Phase A: Iterative pulse period estimation <- $FFT$ -> $FilA$
• Phase B: Generic cardic signal filter & RWAMF -> $FilB$ <- $FilC$
• Phase C: Vital information extraction -> $bpm$, $s_{BP}$, ...

3. Main Outcomes

• Pulse repetition interval, heartbeat rate, abnormalities detection
• Peaks identification, Blood Pressure Waveform
• Respiratory monitoring
• Extracted vital feature parameters could be studied for biometric authentication and encryption

Warning

Spectrum-based HR estimation in complex signal environments remains a significant open challenge in the field. Please be aware that no single algorithm currently offers universal applicability across all scenarios. The performance of the algorithm heavily relies on the specific conditions during signal collection, especially antenna pointing. We are continuously researching solutions to improve this phase and welcome community feedback for further optimization.

Phase Results

Signal separation

• Extract breathing signal $s_{b}$ with FIR linear-phase filter
• Heartbeat signal -> $s_{h}=s_{vital}-s_{b}$

---

RWAMF design

• Calculate the average waveform based on the extracted cardiac signal as the tmplate signal of the filter

---

Cardiac pulse identification

• Main function: findpeaks in MATLAB

---

Blood pressure waveform reproduction

Overall Result

Case 1: with oximeter

---

Case 2: with ECG signal

FYI

Citation

@ARTICLE{wu2025vs,
  author={Wu, Ruochen and Miro, Laura and Aguasca, Albert and Najar, Montse and Broquetas, Antoni},
  journal={IEEE Transactions on Mobile Computing}, 
  title={Robust Biometric Information Sensing With mmWave Radar System-on-Chip}, 
  year={2025},
  volume={25},
  number={5},
  pages={6914-6928},
  doi={10.1109/TMC.2025.3640267}
}

---

Contribution

⭐️ Thank you for your interest! ⭐️