Background
Wearable health monitoring systems have long been hindered by the limitations of conventional von Neumann architectures, which suffer from high energy consumption, computational latency, and bulky hardware when processing multi-modal physiological signals such as electroencephalography (EEG) and snore audio. To enable truly seamless, continuous, and unobtrusive sleep monitoring, brain-inspired computing is urgently needed. Neuromorphic electronics, particularly physical reservoir computing (PRC), offers a promising route by leveraging the intrinsic dynamics of hardware devices to perform analog signal processing without extensive training.
Research Progress
To address these challenges, a research team led by Prof. Tianyu Wang and Jialin Meng from the School of Integrated Circuits and State Key Laboratory of Crystal Materials at Shandong University has developed the world’s first fiber-shaped memristor based on molybdenum disulfide (MoS₂) quantum dots for physical reservoir computing in wearable sleep monitoring. The device features an Ag/MoS₂ quantum dot/Ag coaxial structure, exhibiting excellent mechanical flexibility and stable pulse-programmable conductance suitable for textile integration.
The core innovation lies in using the memristor’s native nonlinear dynamics as a physical reservoir—enabling direct, in-situ mapping of raw EEG waveforms and snore audio into high-dimensional feature spaces without digital preprocessing. Operating at ±1 V and sub-nanoampere currents, the system achieves pico-watt-level power consumption, ideal for all-night wear. In real-world validation, it achieved 94.8% accuracy in snore event detection, 95.4% in sleep stage classification, and 93.5% in multi-modal fusion tasks. Notably, the team replaced the conventional linear readout layer with a lightweight convolutional neural network (CNN),
which significantly improved robustness against environmental noise (e.g., rain, traffic) while accelerating inference by 6× compared to standard deep learning pipelines.
Beyond sleep monitoring, the platform demonstrated versatility in handwritten character recognition and image compression, confirming its potential as a general-purpose analog signal processor woven directly into fabrics.
Future Prospects
Looking ahead, this fiber memristor technology could be seamlessly integrated into everyday textiles to enable passive, continuous, and low-cost health monitoring in domestic settings. Its ultra-low power profile and compatibility with large-scale weaving processes make it highly scalable for consumer and clinical use. Applications extend to elderly care (e.g., detecting nocturnal seizures or sleep disorders), mental health tracking via sleep quality metrics, and even human–machine interfaces in smart homes. Moreover, the success of this PRC-on-fiber approach paves the way for next-generation intelligent e-textiles that merge sensing, computing, and communication within a single thread—ushering in a new era of “invisible” AI for personalized healthcare.

The complete study is accessible via DOI:10.34133/research.0870