Invisible Preventive Health Awareness: A Context-Aware Smart Belt for Posture and Physiological Sensing

Abstract

Musculoskeletal disorders and posture-related conditions are increasingly recognized as major public health concerns in modern sedentary societies. Belt over-tightening is a common but rarely monitored daily habit that has been associated with increased abdominal compression, altered pelvic mechanics, and posture-related discomfort. Despite growing evidence of such issues, current self-assessment strategies remain largely subjective, and many existing systems lack the unobtrusive and socially acceptable design necessary for long-term everyday adoption. This paper presents T-sense, a context-aware wearable system designed to support everyday awareness of belt tension, posture, and physiological rhythms through sensing embedded in a conventional accessory. The proposed solution integrates a load cell in the lumbar segment of the belt and couples strain-based tension measurement with an embedded neural network running on a microcontroller. Unlike pressure-monitoring solutions that interpret tension in isolation, T-sense contextualizes measurements according to user activity, such as Walking, Standing, Sitting, and Slouching. Moreover, the same sensing element is used to explore the feasibility of deriving physiological indicators such as walking cadence and respiratory rhythm. A feasibility evaluation on approximately 23,000 samples collected from 8 participants achieved 92.92% within-subject activity classification accuracy after user-specific calibration, while a leave-one-participant-out evaluation achieved 75.52% accuracy in a cold-start scenario on unseen users. These results suggest that T-sense can provide useful initial awareness for new users while benefiting from user-specific calibration during continued use. Preliminary signal analysis further indicates that walking cadence and respiratory rhythm can be estimated from the belt-tension signal under controlled conditions.