A Virtual Reality Dataset to Support Hand Action Observation in Rehabilitation and Motor Learning Studies

Abstract

The recovery of motor function is increasingly understood as a process influenced not only by physical training but also by perceptual and cognitive strategies. Action Observation Treatment (AOT), a neurorehabilitation approach in which patients observe goal-directed motor actions before executing them, has demonstrated clinical benefits; however, its wider implementation is hindered by a lack of standardized procedures. We present an open-access dataset of 33 upper limb gestures specifically developed to support the administration of Virtual Reality-based AOT (VR-AOT). The gestures were selected in collaboration with expert physiotherapists to ensure clinical relevance, and are provided as motion capture recordings along with Unity-based 3D animations embedded in configurable virtual scenes. The dataset is designed for flexibility, allowing users to modify parameters such as viewpoint, laterality, and repetition count. Technical validation confirms its usability and therapeutic applicability across multiple clinical and research contexts. This dataset offers a standardized yet customizable resource for developing and comparing VR-AOT protocols, with potential applications in neurorehabilitation and motor learning research.