Human Hand Kinematic Models for Sensor Glove-Based HMI: A Review
TL;DR
The reviewed literature indicates that rigid-body models remain the most widely used approach because of their computational efficiency and compatibility with wearable sensing systems, whereas musculoskeletal models provide greater anatomical realism at higher computational cost.
Abstract
Abstract The human hand contains 27 bones, 36 articulations, 39 active muscles, and is commonly represented by approximately 20–27 functional degrees of freedom, depending on the adopted kinematic model. Accurate reconstruction of hand motion is essential for sensor glove-based human–machine interfaces (HMIs) used in robotics, rehabilitation, teleoperation, virtual and augmented reality, prosthetics, and gesture-based interaction. This review analyzes human hand kinematic models and their integration with sensor glove technologies for motion capture and hand pose reconstruction. Rigid-body, musculoskeletal, data-driven, and hybrid modeling approaches are compared with respect to anatomical fidelity, computational complexity, and suitability for real-time applications. The review also examines sensing technologies, including flex sensors, inertial measurement units, optical tracking systems, and multimodal sensing architectures, together with calibration and sensor fusion methods. The reviewed literature indicates that rigid-body models remain the most widely used approach because of their computational efficiency and compatibility with wearable sensing systems, whereas musculoskeletal models provide greater anatomical realism at higher computational cost. The review identifies hybrid modeling, standardized calibration procedures, multimodal sensor fusion, and benchmark datasets for hand motion reconstruction as key directions for future research in sensor glove-based HMIs.