Computer-Vision-Assisted Palm Rehabilitation With Supervised Learning.

Area of Science:

• Biomedical Engineering
• Rehabilitation Science
• Computer Vision

Background:

• Computer-assisted interfaces are increasingly used in physical rehabilitation.
• Contactless methods offer advantages for patient comfort and accessibility.
• Palm and finger rehabilitation requires precise tracking of 3D movements.

Purpose of the Study:

• To propose a computer-vision-assisted contactless methodology for palm and finger rehabilitation.
• To develop a Unity3D-based interface for analyzing rehabilitation movements and providing online feedback.
• To evaluate the effectiveness of different machine learning models for gesture classification in rehabilitation.

Main Methods:

• Utilized a Leap Motion controller to capture 3D palm and finger movements.
• Developed a rehabilitation interface using the Unity3D platform.
• Employed Linear Discriminant Analysis (DA) and Support Vector Machines (SVM) for isolated gesture classification.
• Applied discrete Hidden Markov Models (HMM) for gesture sequence classification.
• Compared HMM performance with Conditional Random Fields (CRF) based techniques.

Main Results:

• DA and SVM showed similar performance in isolated gesture recognition.
• HMM and CRF demonstrated comparable results in classifying gesture sequences.
• The system successfully analyzed intermediate rehabilitation steps and provided online user feedback.
• Experimental validation with healthy volunteers confirmed the system's capabilities.

Conclusions:

• The proposed computer-vision-assisted system facilitates effective contactless palm and finger rehabilitation.
• The system enables analysis and feedback without requiring expert supervision, supporting home-based therapy.
• Machine learning models like DA, SVM, and HMM are suitable for gesture recognition in rehabilitation contexts.