Identifying Key Hand Joints in Grasping Tasks for Wearable Applications*

Wearable robots have been used for the rehabilitation and assistance of people with reduced hand mobility, such as stroke and spinal cord injury survivors. Most of these devices rely on hand kinematic information to close the control loops. However, the biomechanical complexity of the hand makes kinematics data acquisition and processing an open challenge. To tackle this, we present a method to identify the most informative joints of the hand with respect to hand global posture. We processed kinematics data belonging to two public databases, gathered during the execution of grasping tasks of activities of daily living (ADLs). To carry the analysis, we employed principal component analysis (PCA), standard deviation (SD), and covariance matrix (CvM). Our results suggest that flexion/extension motions of the metacarpal joint of the ring finger bring the most informative content during grasping tasks, whilst abduction/adduction motions are the least relevant. The proposed method represents a simple yet powerful tool for the strategic identification of the most informative joints involved in grasping tasks, which can be used for minimizing the number of sensors and optimizing their placement in wearable hand robots, by generalizing the available information and reducing kinematic redundancy.