Virtual hand rehabilitation training system based on neural network-assisted surface electromyography feature extraction and analysis

Scritto il 30/09/2026
da Rixi Huang

J Hand Ther. 2026 Sep 30:S0894-1130(26)00077-3. doi: 10.1016/j.jht.2026.06.019. Online ahead of print.

ABSTRACT

BACKGROUND: Traditional hand rehabilitation relies heavily on professional therapists, which brings great challenges and limitations to patients with hand dysfunction, making convenient and flexible home rehabilitation difficult to achieve.

PURPOSE: This study aims to develop an innovative virtual hand rehabilitation system to improve patients' finger motor coordination and provide an effective, convenient home rehabilitation solution.

STUDY DESIGN: This study adopted a technical development and preliminary verification design. A targeted virtual rehabilitation system was developed and preliminarily validated through patient participation trials.

METHODS: Based on the Unity platform and C# scripting, the system integrates VR technology, sEMG real-time signal monitoring and LSTM neural networks. Three virtual rehabilitation game scenarios were constructed to realize interactive training and accurate gesture recognition.

RESULTS: The system achieves over 90% accuracy in recognizing eight complex hand gestures. Ten patients with hand dysfunction participated in the test, and all feedback confirmed improved rehabilitation engagement, safety and convenience.

CONCLUSIONS: The proposed system has reliable recognition performance and good user experience. It offers a novel and feasible approach for home-based fine motor rehabilitation of hand dysfunction patients with broad application potential.

PMID:42816195 | DOI:10.1016/j.jht.2026.06.019