Opto-Electronic Advances, Volume. 8, Issue 5, 240254(2025)

Wearable photonic smart wristband for cardiorespiratory function assessment and biometric identification

Wenbo Li, Yukun Long, Yingyin Yan, Kun Xiao, Zhuo Wang, Di Zheng, Arnaldo Leal-Junior, Santosh Kumar, Beatriz Ortega, Carlos Marques, Xiaoli Li, and Rui Min
Figures & Tables(10)
Functional architecture of the proposed wristband, including mainly five primary components: the ALL-PSU sensing unit, signal extraction and preprocessing, machine learning models, and PSD analysis.
(a) The conceptual figure of the PSP fiber adapter. (b) Pressure simulation (left) alongside the actual configuration of the PSP fiber adapter (right), demonstrating the adapter's response to pressure. (c) Exploded schematic of the practical All-PSU, including 3D molds, DSCs, PSP fiber adapter, and a glycerol-filled capsule. (d) Structural diagram of the All-PSU, the entire process of All-PSU is divided into three parts: the PSP fiber adapter fabrication, the PSP-DSC and liquid capsule base fabrication, and the All-PSU construction and coating.
(a) Method to extract clean and accurate pulse wave signals as well as respiratory signals. (b) PSD analysis method utilized to monitor RR and HR. (c) Feature extraction process and the seven distinct features extracted from the pulse wave signals, which are utilized for both BP estimation and biometric ID (H represents the subject height). (d) Architecture of the GRU model developed for BP estimation. (e) Structure of the RF classifier implemented for biometric ID.
(a) Schematic diagram of the PSP-DSC and All-PSU performance test experiments. (b) Transmitted power of PSP-DSC under varying pressures, with pressure increased by 10 N each time. The line plot illustrates the mean value of the transmitted light power derived from the three repeated experiments. The gray area surrounding the line plot represents the data distribution across these trials. (This applies similarly to subfigure (d) and (f)). (c) Transmitted power of three PSP-DSC samples under varying pressures, with pressure increased by 10 N each time. (d) Transmitted power of PSP-DSC under different bending angles, with angles increased by 10° each time. (e) Transmitted power of three PSP-DSC samples under different bending angles, with angles increased by 10° each time. (f) Transmitted light power of PSP-DSC under tensile strains, with deformation increased by 10% every 30 seconds. (g) Transient pulse response time results (response time: 6 ms, recovery time: 12 ms).
(a) The dynamic response of the All-PSU. Evaluated across a range of tensile velocities on stepper motors. (b) Performance assessment of the All-PSU through cyclic tensile strain testing, subjected to a tensile strain of 50% over 1,000 cycles. (c) Waterproofing of the All-PSU confirmed by immersion in water at room temperature (23 °C) for 24 hours, with response intensity recorded every 2 hours. (d) Response intensity of the All-PSU placed in water at various temperature. (e) Pulse wave signal monitoring at different wrist locations, with the star shape indicating the position of the All-PSU.
(a) Diagram and processing schematic of RR and HR monitoring system based on PSD analysis. The experiment was mainly divided into three stages: four subjects monitored their RR and HR under relaxed conditions, another continuously monitored his RR and HR from 14:00–18:00, and four subjects monitored their RR and HR under different states (squat. burpee and high knee). Signal processing is mainly divided into signal acquisition, signal noise reduction, signal extraction, frequency domain conversion, and RR and HR calculation. (b) Line box plot of the four subject's results of RR and HR for four subjects. (c) Violin plot of the four subjects measured and predicted RR results, and (d) Violin plot of the four subjects measured and predicted HR results. NS: indicates that there is no significant difference between the results of measured and predicted, indicating that the results of predicted HR, according to the proposed system, are accurate for different subjects.
(a) RR and HR monitoring results at 14:00–18:00. (b) Violin plot comparing measured and reference RR results during 14:00–18:00, showing no significant difference and suggesting accurate continuous RR measurement. (c) Violin plot comparing measured and reference HR results during 14:00–18:00, showing no significant difference and suggesting accurate continuous HR measurement. (d) Scatter plot of RR and HR monitoring results of four subjects in squatting, burpee, and high knee. (e) Line box plot of RR and HR monitoring results of subjects in different statuses. (f) Correlation result plot between predicted and measured RR in different exercises. (g) Correlation result plots between the predicted and measured HR in different exercise states. (h) MAPE plot of RR predictions for subjects in different exercise states. (i) MAPE Plot of HR predictions of four subjects in different exercise states.
(a) Schematic diagram of three pivotal phases and BP estimation model processing the whole BP monitoring process is divided into six subjects to continuously monitor BP for one week to establish a model, another subject monitored BP changes in the morning, afternoon, and evening for one day, and another subject monitored BP fluctuations in different states (resting squat, burpee), the main model processing is divided into signal acquisition, feature extraction, and machine learning model estimation. (b) Correlation analysis graph between SBP measured results and reference results. (c) Correlation analysis plot of DBP measured results and reference results. (d) Bland-Altman plots of SBP, and (e) Bland-Altman plots of DBP.
(a) Line graph of BP measurement results in the morning, afternoon, and evening. (b) Line box plots of BP measured results in the morning, afternoon, and evening. (c) Correlation analysis of SBP measured and reference results. (d) Correlation analysis of DBP measured and reference results. (e) Bland-Altman plot of SBP in the morning, afternoon, and evening. (f) Bland-Altman plot of DBP in the morning, afternoon, and evening. (g) Line graph of BP measurement results at rest, squat, and jump. (h) Line box plot of BP measurements at rest, squat, and burpee. (i) Correlation analysis of SBP measured and reference results under different exercise statuses. (j) Correlation analysis of DBP measured and reference results under different exercise states. (k) Bland-Altman plot of SBP at rest, squat, and burpee. (l) Bland-Altman plot of DBP at rest, squat, and burpee.
(a) Schematic diagram of the biometric identification process, which mainly includes pulse signal collection, pulse feature extraction, personal information database establishment, RF algorithm decision-making, and identification results. (b) The weighting pie chart of the seven features in the biometric identification process. (c) Confusion matrix for biometric identification results. (d) Biometric identification accuracy of different types of subjects and total prediction accuracy of three subjects. (e) The customized APP, a personal information interface, a week historical data interface, and a real-time measurement interface, and (f) the smartwatch worn by a 24-year-old male.
Tools

Get Citation

Copy Citation Text

Wenbo Li, Yukun Long, Yingyin Yan, Kun Xiao, Zhuo Wang, Di Zheng, Arnaldo Leal-Junior, Santosh Kumar, Beatriz Ortega, Carlos Marques, Xiaoli Li, Rui Min. Wearable photonic smart wristband for cardiorespiratory function assessment and biometric identification[J]. Opto-Electronic Advances, 2025, 8(5): 240254

Download Citation

EndNote(RIS)BibTexPlain Text
Save article for my favorites
Paper Information

Received: Oct. 26, 2024

Accepted: Mar. 3, 2025

Published Online: Aug. 5, 2025

The Author Email:

DOI:10.29026/oea.2025.240254

Topics