Researchers from Xiamen University and North University of China have introduced a new wearable device that uses wrist pulse impulses to directly assess respiration patterns. The study, which was published in Microsystems & Nanoengineering on May 16th, 2025, describes a small, artificial intelligence (AI)-powered gadget that combines a deep neural network with a flexible pressure sensor.
Schematic illustration of the flexible sensor system. a The flexible sensor is mounted on the radial tuberosity of the wrist to capture the pulse signal, which is then processed to ascertain real-time respiratory status using the ResNet-BiLSTM model. b A schematic layout of the system is provided, showcasing the integration of the pulse detection sensor and the flexible circuit for data acquisition and processing. c Detailed diagram of the data acquisition and processing flexible circuit. Image Credit: Chinese Academy of Sciences
Millions of people worldwide suffer from chronic respiratory diseases, which need ongoing observation to control symptoms and avoid consequences. However, several of the current technologies, such as nasal prongs or chest straps, might be invasive and inappropriate for regular, prolonged usage.
Indirect indications like heart rate or pulse waveforms have been investigated in recent methods; they provide more comfort but frequently come at the expense of accuracy.
The intricacy of human respiration makes traditional signal categorization techniques inadequate. Owing to these difficulties, there is an increasing need for devices that can extract high-fidelity respiratory data over long periods of time while simultaneously being comfortable to wear. These difficulties highlight the urgent need to create sophisticated methods for accurate, long-term respiratory monitoring.
The center of the system is a flexible pressure sensor 300 μm thick and modeled after the structure of a human fingertip. The sensor is printed on a thermoplastic polyurethane (TPU) substrate and can identify minute variations in pulse waves caused by respiration.
A hybrid Residual Network – Bidirectional Long Short-Term Memory (ResNet-BiLSTM) neural network processes these signals, which are classified as respiration-induced amplitude variation (RIAV), respiration-induced fluctuation in ventricular filling (RIFV), and respiration-induced variation in baseline (RIIV).
The signals are sent via Bluetooth to a mobile application. With an impressive 99.5% classification accuracy, this deep learning model captures the temporal and spatial dynamics of respiratory patterns, distinguishing between slow, normal, rapid, and simulated breathing phases.
The device’s ultra-lightweight construction (just 9 grams), skin-conforming shape, and long-term mechanical stability make it comfortable to wear for hours. Thirteen human volunteers participated in the testing, which featured machine-simulated breathing and demonstrated the sensor's stability across actual and artificial respiratory settings.
The device simplifies setup and improves use by minimizing dependency on chest positioning or airflow proximity, making it suitable for everyday usage.
Our mission was to bridge the gap between high-precision monitoring and wearable comfort. We’ve shown that you can track respiration with clinical accuracy—without putting anything on your chest or face. This could be a game-changer in how we approach remote health monitoring, especially for patients who need round-the-clock care.
Dr. Libo Gao, Study Senior Author and Professor (Associate), Xiamen University
The device's ability to provide accurate respiratory insights in a pleasant, wearable format offers new opportunities in chronic disease management, eldercare, and telemedicine. Its seamless interaction with mobile platforms enables real-time alerts and long-term data logging, critical for early intervention in diseases such as COPD or sleep apnea.
Beyond healthcare, this breakthrough might aid athletes, astronauts, and high-altitude workers by providing continuous respiration tracking in dynamic conditions. As wearable technology advances, this wrist-worn device might become a cornerstone in the future of individualized respiratory health monitoring.
The National Key Research and Development Program of China (2023YFB3208600), the National Natural Science Foundation of China (No. 62274140), Key Program of the National Natural Science Foundation of China (62433017), and the Science and Technology on Vacuum Technology and Physics Laboratory Fund (HTKJ2023KL510008), the Fundamental Research Funds for the Central Universities (20720230030), the Xiaomi Young Talents Program/Xiaomi Foundation, Shenzhen Science and Technology Program (JCYJ20230807091401003) supported the study.
Journal Reference:
Zhang, X., et al. (2025) Direct extraction of respiratory information from pulse waves using a finger-inspired flexible pressure sensor system. Microsystems & Nanoengineering. doi.org/10.1038/s41378-025-00924-4.