Environmental Features Recognition for Lower Limb Prostheses Toward Predictive Walking.

Area of Science:

• Biomedical Engineering
• Robotics
• Computer Vision

Background:

• Lower limb prostheses require advanced control systems to adapt to diverse environments.
• Accurate recognition of environmental features is crucial for seamless prosthetic limb adaptation.

Purpose of the Study:

• To develop and validate a robust environmental features recognition system (EFRS) for lower limb prosthesis control.
• To enhance prosthetic functionality by predicting amputee locomotion modes and estimating environmental characteristics.

Main Methods:

• Integration of a depth sensor and inertial measurement unit to stabilize 3D environmental point clouds.
• 2D point cloud extraction and classification using a neural network for environmental feature estimation (slope, stair dimensions).
• Validation through simulation, indoor, and outdoor experiments with healthy subjects and amputees.

Main Results:

• High classification accuracy for common environments (up to 99.3% indoors, 98.5% outdoors for amputees).
• Locomotion modes predicted over 0.6 seconds in advance.
• Low environmental feature estimation errors (under 5% indoors, under 10% outdoors).
• Rapid processing time of less than 0.023 seconds.

Conclusions:

• The EFRS demonstrates robustness and significant potential for improving lower limb prosthesis control.
• The system accurately recognizes diverse environments and predicts user intent, enhancing mobility for amputees.