UCLA

Respiratory monitoring is crucial because it provides key insights into a person's health and physiological condition. Conventional respiratory sensing is significantly challenged by the presence of water vapor in exhaled breath. A on-mask magnetoelastic sensor network is developed, which combines mechanically informed learning with an ultralight, intrinsically waterproof architecture to achieve continuous, long-term respiratory monitoring and real-time, high-fidelity signal acquisition. Leveraging the giant magnetoelastic effect, each sensing unit is miniaturized to just 3.2 grams, which markedly enhances its sensitivity to airflow-induced mechanical fluctuations during respiration, while also ensuring sufficient wearing comfort for daily use. Beyond mechanical compliance, the system achieves a signal-to-noise ratio exceeding 35 dB and a rapid response time of 80 ms under optimal conditions, and it can reliably transduce the fluid dynamics generated during respiration in the mouth–mask microenvironment into high-fidelity electrical signals for continuous respiratory monitoring. With the aid of machine learning, the on-mask magnetoelastic sensor network achieves respiration pattern recognition with a classification accuracy of up to 94.03%. Furthermore, a user-friendly, custom-designed mobile application has been developed to process respiratory signals, enabling real-time, data-driven diagnosis and seamless one-click health data sharing with clinicians. This machine-learning-enhanced sensor network paves the way for personalized respiratory management in the Internet of Things era.