Performance Evaluation of Leap Motion Controller for Hand Motion Assessment

The use of markerless depth cameras for hand kinematic reconstruction is gaining interest due to their ability to enable more natural and unconstrained motion analysis. The leap motion controller (LMC) is a compact and widely adopted example, equipped with an integrated skeleton-tracking algorithm that estimates 3-D joint positions without the need for reflective markers. Several studies in the literature have assessed the accuracy of the LMC in reconstructing hand kinematics; however, they have generally overlooked the following critical aspects: 1) the use of kinematic protocols capable of capturing metacarpophalangeal (MCP) joint abduction–adduction (A/A); 2) the presence of static and dynamic offsets that can degrade angle estimation accuracy; and 3) the evaluation of additional kinematic features such as intrafinger couplings. In this article, an offset correction strategy is proposed to improve the accuracy of LMC in reconstructing hand joint angles during flexion–extension (F/E) and A/A movements. The proposed approach was evaluated through a comparative analysis of the joint angles extracted using the LMC and a marker-based optoelectronic system (MOS). Ten healthy subjects were recruited to collect joint angle data during F/E and A/A movements. The results demonstrated significant improvements in angle estimation accuracy: in 95% of cases, the angular error was below 0.17 rad, and in 85% of cases, the Spearman correlation coefficients exceeded 0.7. Nevertheless, the Bland–Altman analysis indicated that the LMC and MOS measurements are not yet interchangeable in applications where high accuracy is required, such as clinical evaluation and hand prosthesis design.