A new study reveals a smart garment that could change the way sleep disorders are monitored. Equipped with a deep learning-enabled strain sensor array, this wearable technology offers continuous and precise tracking of sleep-related conditions, addressing key limitations of traditional monitoring systems. Researchers say the innovation could help to lay the groundwork for a more practical and user-friendly approach to sleep health tracking.
Study: A deep learning-enabled smart garment for accurate and versatile monitoring of sleep conditions in daily life. Image Credit: Tero Vesalainen/Shutterstock.com
Background
Current sleep monitoring systems often rely on bulky equipment, including wired sensors and rigid wearables, which can be uncomfortable for users and impractical for regular use. These setups frequently require direct skin contact, making them intrusive and potentially leading to inconsistent readings due to movement or improper positioning. Additionally, they are susceptible to motion artifacts and external noise, which compromise data accuracy.
Sleep disturbances—such as mouth breathing, snoring, and sleep apnea—are widespread, affecting millions of individuals and contributing to chronic health issues, including cardiovascular disease, diabetes, and cognitive decline. Despite the clear need for improved sleep monitoring, existing solutions remain inconvenient and inaccessible for many people. There is an urgent demand for a user-friendly, nonintrusive, and highly accurate technology that can monitor sleep patterns effectively in real-world settings.
The Study
To address these issues, the researchers developed a washable, skin-compatible smart garment with a printed textile-based strain sensor array. Unlike conventional sensors, this design works effectively with minimal contact, eliminating the need for precise positioning or skin preparation. The sensor detects subtle skin movements, measuring strain signals from 0.1 % to 10 % with a gauge factor of 100. A specially designed printed pattern helps isolate relevant signals from external motion interference, while a reversible starching treatment ensures consistent performance across different manufacturing batches.
To analyze the collected data, the system integrates advanced deep learning techniques, including explainable AI and transfer learning. The deep learning model, SleepNet, classifies six different sleep states with an impressive accuracy. The garment captures signals from sleep-related movements and vibrations, analyzing them in real time to detect nasal and mouth breathing, snoring, bruxism, and both central (CSA) and obstructive sleep apnea (OSA). SleepNet is also highly adaptable, achieving high accuracy even with minimal training data for new users.
Results and Implications
The smart garment demonstrated remarkable accuracy in capturing and analyzing sleep-related signals. Utilizing a multichannel graphene textile strain sensor array, it effectively records vibrations from extrinsic laryngeal muscles, which correspond with different sleep patterns. Its strain isolation technique reduces motion artifacts, improving data reliability. Testing confirmed the garment's ability to distinguish breathing patterns and detect signs of sleep disturbances with high precision.
One of the most notable findings is the garment's 98.6 % accuracy rate, outperforming existing wearable sleep monitoring devices. Beyond accuracy, the garment is designed for scalable manufacturing, allowing flexibility in production without sacrificing performance. Since it does not require precise positioning, it is more practical and comfortable for everyday use.
The study suggests that smart garments not only surpass current sleep monitoring technologies but also have strong potential for widespread consumer adoption. The system’s adaptability is another highlight, achieving up to 95 % classification accuracy for new users after minimal exposure to training data. This positions the smart garment as a promising innovation in modern sleep health technology.
Conclusion
This study represents a major step forward in wearable sleep monitoring technology. By integrating a textile-based strain sensor array with deep learning, researchers have developed a comfortable, durable, and nonintrusive solution for continuous sleep tracking. Given the challenges of accurately assessing sleep conditions, the smart garment stands out for its high performance, adaptability, and ease of use.
With applications in both personal health and consumer electronics, this research highlights the growing role of smart textiles in sleep health. As technology advances, further improvements could enhance the detection and management of sleep disorders, making wearable devices a key tool in everyday well-being.
Journal Reference
Tang C., Yi W., et al. (2025). A deep learning-enabled smart garment for accurate and versatile monitoring of sleep conditions in daily life. Proceedings of the National Academy of Sciences, 122(7), e2420498122. DOI: 10.1073/pnas.2420498122, https://www.pnas.org/doi/10.1073/pnas.2420498122