The design of this group combines rich immersive VR scenes and specific interactive tasks. Specifically, there were two VR scenes: (1) Natural scenery: The subjects were on a quiet forest path, surrounded by tall trees, birdsong, and the sound of gurgling water. In this scene, the program sets the virtual hand on the hemiplegic side to perform the task of grasping the sphericalfruit on the tree. Each time the fruit is grasped for 3 seconds, and then released for 3 seconds; (2) City park: The subject was located in a city park surrounded by the sounds of children playing and the background sounds of the natural environment. In this scenario, the program sets the virtual hand on the hemiplegic side to carry out the task of grasping a ping-pong ball in the park, holding it for 3 seconds at a time and then releasing it for 3 seconds. To improve the actual participation, a real object the same as the virtual object will be provided in the real world, and subjects will be instructed to use

The VR scene in this group focuses on basic motor interaction, and the virtual background is the same as the control group. The program will set the virtual hand on the hemiplegic side to complete specific interaction tasks, such as grasping the virtual ball, holding the virtual ball for 3 seconds each time, and then releasing it for 3 seconds. At the same time, a real sphere of the same size and color as the virtual ball will be provided in the real world, and the subject will be instructed to perform the same movement simultaneously as much as possible with the real hand on the hemiplegic side.

subjects were asked to wear an HTC Vive HMD and enter a simple virtual testing environment. The environment has no complex narrative or visual effects, nor does it emphasize specific interaction tasks. It only includes a realistic virtual hand model of the hemiplegic side and the same virtual background as the real experiment, both are presented from a first-person perspective. The program will set the virtual hand on the hemiplegic side to perform simple and repetitive grasping movements. Each fist hold lasts for 3 seconds and then releases for 3 seconds. At the same time, the subject will be instructed to perform the same movement simultaneously with the real hand on the hemiplegic side as much as possible.

This group of subjects received immersive and interactive VR training, which combined rich immersive VR scenes with specific interactive tasks, including two VR scenes. Task selection was based on the subjects' upper limb functional performance, and experienced therapists provided guidance and monitoring, and adjusted the tasks when necessary.

The Fugl-Meyer Assessment of the Upper Extremity (FMA-UE) covers all aspects of upper limb function, such as movement, coordination, and reflex actions of the shoulder, elbow, forearm, wrist, and hand, with a total of 33 assessment items. Patients were assessed according to a specific scale with a total score of 66, with higher scores indicating better recovery of upper limb function.

The sense of ownership (SOO) questionnair investigates a number of aspects including sense of body ownership, sense of position in the real or virtual hand, haptic experience, body similarity, and body transformation. Scores are allocated on a 7-point Likert scale, from "Strongly disagree (-3 points)" to "Strongly agree (3 points)", depending on how much the patient agrees with each statement. The mean score of all questions is calculated as the final SOO questionnaire score, and the higher the score, the stronger the SOO.