A Subvision System for Enhancing the Environmental Adaptability of the Powered Transfemoral Prosthesis

Area of Science:

• Biomedical Engineering
• Robotics
• Human-Computer Interaction

Background:

• Human locomotion relies heavily on visual information, which is challenging for amputees using prostheses.
• Current prosthetic technology often lacks direct neural signal transmission for intuitive control.
• Transfemoral amputees face difficulties in navigating complex environments with conventional prosthetic limbs.

Purpose of the Study:

• To develop and evaluate a subvision system for augmenting human-prosthesis interaction.
• To enable predictive control of powered prostheses by perceiving environmental cues.
• To reconstruct a complete vision-locomotion loop for transfemoral amputees.

Main Methods:

• Utilizing deep learning for environmental perception and motion intent estimation.
• Classifying common static terrains (e.g., level ground, stairs, ramps) with high accuracy.
• Integrating the subvision system with a powered prosthesis locomotion control system.

Main Results:

• The subvision system achieved 98% accuracy in classifying terrains and estimating motion intents.
• Powered prostheses demonstrated enhanced natural, nonrhythmic locomotion, including mode switching and obstacle crossing.
• The system successfully recognized dynamic objects and facilitated agile obstacle-avoidance reflexes.

Conclusions:

• The subvision system effectively reconstructs a complete vision-locomotion loop for transfemoral amputees.
• This integration significantly enhances the environmental adaptability and mobility of amputees using powered prostheses.
• The developed system represents a significant advancement in prosthetic limb control and human-robot interaction.