From Lab to Bed: Robust and Adaptive Non-Contact IoT Sleep Monitoring for Smart Health

Poor sleep quality has far-reaching consequences, including increased incidence of psychological disorders, cognitive impairment, and physical illness, which also place significant strain on caregivers. The current gold standard for sleep assessment, polysomnography, suffers from limitations such as complexity, high cost, interfere with participants' natural sleep behavior and the need for specialized laboratory settings and trained professionals. To address these challenges, we propose a novel contactless sensing, lightweight real-time IoT system designed specifically for monitoring vital signs and posture under blankets, eliminating the need for physical contact or wearable sensors, ensuring comfort and minimizes patient anxiety. The system utilizes an incremental learning approach to adapt to out-of-distribution datasets during long-term deployment, enhancing robustness. Our research contributes to the advancement of non-invasive sleep monitoring technology, providing a valuable tool for assessing sleep quality without disrupting participants' natural sleep patterns.