Smart Wearable Sensors for Healthcare and Inclusive Communication: From Sarcopenia Risk Assessment to Sign Language Recognition

Abstract

With the increasing global aging population and growing needs for assistive technologies, the development of intelligent wearable systems for human motion analysis and gesture recognition has become crucial for healthcare monitoring and inclusive communication. This dissertation presents three interconnected studies advancing the field of intelligent sensing systems, addressing critical challenges in early sarcopenia detection and sign language translation.

The first study introduces a novel machine learning-based approach for multi-risk-level sarcopenia screening using wearable inertial measurement units (IMUs). While traditional diagnostic methods like Dual-energy X-ray absorptiometry (DEXA) and Bioelectrical impedance analysis (BIA) require specialized equipment and expertise, our approach offers an accessible alternative through sit-to-stand motion analysis. In a study of 53 older adults (65-84 years), we extracted 510 features from four distinct motion phases using strategically placed IMUs. The system achieves 98.32% accuracy in distinguishing healthy from sarcopenia-prone individuals using Support Vector Machine and Multilayer Perceptron algorithms, and 90.44% accuracy in classifying four risk levels using K-Nearest Neighbors, advancing beyond conventional binary classification approaches.

Building on these findings, the second study presents a comprehensive analysis of the Five-Times-Sit-to-Stand Test (5TSTS), extending the methodological framework established in the first study while increasing its clinical depth. By implementing wavelet transform analysis across multiple movement phases (Standing-up, Standing Transition, Sitting-down, and Sitting Transition), we extracted 379 time-frequency domain features per phase. Through XGBoost-based feature selection and SVMSMOTE data augmentation, our optimized model achieves remarkable performance: 99.28% accuracy in binary classification (healthy vs. sarcopenia-prone) and 97.97% accuracy in multi-class risk level classification, demonstrating the effectiveness of temporal-spectral analysis in capturing subtle movement characteristics associated with sarcopenia progression.

The third study shifts from healthcare monitoring to inclusive communication while maintaining the core principles of smart wearable sensor design. Recognizing that traditional sensors are often hard, rigid, and intimidating for users, particularly elderly individuals reluctant to adopt new technologies, this study introduces an innovative crochet-based capacitive sensor system for sign language recognition. By combining centuries-old crochet techniques with modern liquid metal technology, this smart wearable approach creates textile-integrated sensors that users can actively participate in fabricating. The Galinstan-filled silicone thread sensors achieve exceptional mechanical properties with >30% strain range, 98 ms response time, and excellent stability over 800 cycles. The compact hardware implementation (2.4 × 1.2 × 0.8 cm, 1.66g) with ESP32-C6 microcontroller enables multi-channel capacitive measurements with 28-bit resolution. Machine learning integration using Support Vector Machine classification achieves 96.67% accuracy for Chinese Sign Language gesture recognition, demonstrating effective inclusive communication capabilities across multiple anatomical locations.

These three studies form a coherent research progression that demonstrates the evolution of smart wearable sensors from healthcare monitoring to inclusive communication applications. The methodological advancement from rigid IMU-based systems to soft, crochet-based sensors reflects a deeper understanding of user acceptance and the importance of creating wearable technology that feels familiar and accessible. The transition from sarcopenia risk assessment to sign language recognition showcases the versatility and adaptability of smart wearable sensor platforms for diverse applications serving vulnerable populations.

This research represents not just technical achievements but a comprehensive approach to developing smart wearable sensors that can truly make a difference in people's lives. The integration of traditional craftsmanship with modern sensing technology, the development of practical healthcare monitoring solutions, and the advancement of assistive communication technologies all contribute to a vision of inclusive smart wearable systems. Together, these studies establish a foundation for next-generation wearable sensing platforms that promote independent living for older adults and facilitate inclusive communication for the deaf community, demonstrating that effective smart wearable sensors can be both sophisticated and approachable.

Date of Award	28 Aug 2025
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Wen Jung LI (Supervisor)