Fault Signals Reconstruct for Hand Kinematic Measurement Based on Compressed Sensing

The measurement of hand kinematics is typically achieved through various approaches, such as data gloves, multiple cameras and optical trackers. These methods find applications in diverse fields, including 3D modeling, robotics, rehabilitation, and physiotherapy. However, each approach necessitates a significant number of sensors for accurate hand kinematic measurement. Data gloves, in particular, are prone to calibration issues, and they may encounter missing information due to fault sensors. Defects in these sensors can lead to safety and accuracy issues, especially in the context of using hand kinematics to control robots, particularly in remote surgery as shown that hand kinematic reconstruction is significant to enhance the efficiency and accuracy of hand kinematic analysis across various applications. This paper focuses on hand kinematic signals reconstruction from fault sensors using only measured signals, employing a compressed sensing technique. The proposed sparse sensor placement technique aims to reconstruct all fault signals with suitable accuracy by utilizing measured signals. To evaluate the performance of proposed methods, The Percentage Root-Mean-Square Difference (PRD) used to compare reconstructed signals with ground truth. The results of average PRD shown that the number of faulty sensors directly affects the quality of the reconstruction. As the number of faulty sensors increases, the performance of the reconstruction tends to decrease.