Our group advances musculoskeletal health through the design of intelligent, wearable, and biomechanics-informed systems that enable accurate monitoring and personalized intervention in natural, everyday environments. We focus on transforming traditionally laboratory-bound biomechanical measurements—such as joint loading, muscle forces, and movement kinematics—into accessible metrics derived from lightweight sensors and physics-aware modeling. By combining inertial sensing, surface electromyography, and visual body-shape acquisition with hybrid deep learning architectures, our research bridges the gap between clinical-grade biomechanics and real-world usability.

Building on projects such as ACLGuard and KneeGuard, we develop end-to-end frameworks that integrate multi-modal biomechanical information retrieval with principled modeling inspired by inverse dynamics and neuromuscular control. These systems are designed for high-risk movement assessment, gait re-training, and long-term disease management, with a strong emphasis on knee osteoarthritis and ACL injury prevention. Through innovations in multi-task learning, cross-modality fusion, and physics-aware feature extraction, our work enables precise, calibration-free tracking of knee joint loading and muscle forces. Ultimately, we aim to deliver personalized, real-time feedback technologies that improve movement quality, reduce injury risks, and support effective rehabilitation across clinical and athletic populations.

Related publications:

• [ACM IMWUT 2025] B. Yang, X. Zhang, X. He, C. Xu, W. Xie, Z. Liang, S. YUNG, Q. Zhang, “ACLGuard: Physics-Aware Knee Loading Monitoring System for Anterior Cruciate Ligament Injury Prevention Training.”

  
  

  Introduction: ACLGuard is a physics-aware wearable monitoring system designed for on-field ACL injury-prevention training. It integrates IMU-based continuous motion sensing with a one-time RGB-D body scan to capture essential body information. By combining hybrid deep-learning models with inverse-dynamics-guided multi-task learning, the system accurately estimates knee adduction moment (KAM) even during high-risk, highly dynamic movements. Experiments with athletes and non-athletes show that ACLGuard achieves markerless-motion-capture-level accuracy while maintaining a simple and field-ready setup.

  
  

• [ACM IMWUT 2024] B. Yang, X. Zhang, J. Zhang, Z. Huang, Q. Lu, J. Zhang, H. Hu, Q. Zhang. “KneeGuard: A Calibration-free Wearable Monitoring System for Knee Osteoarthritis Gait Re-training via Effortless Wearing.” [paper]

  
  

  Introduction: KneeGuard is a calibration-free wearable system for gait re-training in knee osteoarthritis patients. It uses IMUs and a circular sEMG array to collect comprehensive biomechanical information, enabling knee loading and muscle-force estimation without any lab calibration. The system incorporates spatial-aware sensing and a biomechanics-inspired multi-task fusion framework to robustly extract muscle and motion features under effortless wearing. Evaluations on KOA patients and healthy subjects show accurate estimation of KAM and key muscle forces, matching prior methods that require calibration.

  
  

• KneeGuard