"Dataset for VR Box and Block Test Upper-Limb Kinematics Using Markerless Vision"

"This dataset supports research on upper-limb kinematics and motor performance during a virtual reality adaptation of the Box and Block Test (VR-BBT). Thirty healthy participants were recruited (18 male, 12 female; ages 20 to 40) with no known neurological or musculoskeletal conditions. Data collection spanned four weeks, with three sessions per week and one final session in the concluding week, totaling ten sessions per participant. Virtual object grasping was facilitated through an electromyography-based activation protocol, and motion data were captured using a monocular RGB camera processed through MeTRAbs for 3D pose estimation. Seven joint angles were computed from the estimated skeletal joints: shoulder flexion, shoulder adduction, elbow flexion, thorax forward tilt, thorax left tilt, head forward tilt, and head left tilt.The dataset comprises four components: (1) normalized movement trajectories from 0 to 100 percent of each box and block transfer cycle, representing one session of video-derived kinematic data; (2) a static validation dataset comparing model-estimated joint angles against goniometer measurements for accuracy assessment; (3) Intrinsic Motivation Inventory survey responses; and (4) Box and Block Test performance scores across all ten sessions. This dataset is intended to serve as a reference source for VR-BBT upper-limb kinematics research, to inform the design and improvement of virtual reality rehabilitation systems through motivational data, and to enable longitudinal studies examining training plateaus and kinematic adaptation over repeated sessions."