GMH-D: Combining Google MediaPipe and RGB-Depth Cameras for Hand Motor Skills Remote Assessment

Abstract

Impairment in the execution of simple motor tasks involving hands and fingers could hint at a general worsening of health conditions, particularly in the elderly and in people affected by neurological diseases. The deterioration of hand motor function strongly impacts autonomy in daily activities and, consequently, the perceived quality of life. The early detection of alterations in hand motor skills would allow, for example, to promptly activate treatments and mitigate this discomfort. This preliminary study examines an innovative pipeline based on a single RGB-Depth camera and Google MediaPipe Hands, that is suitable for the remote assessment of hand motor skills through simple tasks commonly used in clinical practice. The study includes several phases. First, the quality of hand tracking is evaluated by comparing reconstructed and real hand 3D trajectories. The proposed solution is then tested on a cohort of healthy volunteers to estimate specific kinematic features for each task. Finally, these features are used to train supervised classifiers and distinguish between “normal” and “altered” performance by simulating typical motor behaviour of real impaired subjects. The preliminary results show the ability of the proposed solution to automatically highlight alterations in hand performance, providing an easy-to-use and non-invasive tool suitable for remote monitoring of hand motor skills.