A Powered Prosthetic Hand With Vision System for Enhancing the Anthropopathic Grasp

Area of Science:

• Robotics
• Biomedical Engineering
• Human-Computer Interaction

Background:

• Prosthetic hands need anthropomorphic grasping for better user experience.
• Current brain-computer interface (BCI) and electromyography (EMG) systems lack gesture adaptability and intent recognition.
• Vision systems improve object perception but not dynamic gesture control.

Purpose of the Study:

• To develop a vision-powered prosthetic hand system for natural, dynamic grasping.
• To overcome limitations of existing BCI and EMG-based prosthetic hands.
• To enhance user experience and functional efficiency in prosthetic hand use.

Main Methods:

• Spatial Geometry-based Gesture Mapping (SG-GM) models finger joint angles using polynomial functions based on hand-object distance.
• Motion Trajectory Regression-based Grasping Intent Estimation (MTR-GIE) predicts user intent via wrist trajectory regression and object segmentation.
• Integration of vision systems with novel gesture mapping and intent estimation algorithms.

Main Results:

• High anthropomorphism achieved, with a similarity coefficient R²=0.911 and RMSE=2.47°.
• Rapid grasping execution at 3.07±0.41 seconds.
• Robust success rates: 95.43% for single objects and 88.75% for multi-object scenarios.
• MTR-GIE demonstrated 94.35% intent estimation accuracy in multi-object environments.

Conclusions:

• The proposed vision-powered system enables dynamic gesture synthesis for prosthetic hands.
• This approach eliminates the need for invasive sensors like BCI and EMG.
• The system significantly advances the real-world usability and anthropomorphism of prosthetic hands.