Related Experiment Video
Updated: Apr 16, 2026
11:16
Engineering Platform and Experimental Protocol for Design and Evaluation of a Neurally-controlled Powered Transfemoral Prosthesis
Published on: July 22, 2014
16.8K
A Classification Method for User-Independent Intent Recognition for Transfemoral Amputees Using Powered Lower Limb
Summary
User-independent intent recognition systems for powered lower limb prostheses reduce training burden. Mode-specific classification significantly improves transition accuracy for transfemoral amputees, enhancing mobility.
Area of Science:
- Biomedical Engineering
- Rehabilitation Technology
- Human-Machine Interfaces
Background:
- Powered lower limb prostheses aim to restore mobility for transfemoral amputees.
- Intent recognition systems enable seamless transitions between locomotion modes.
- Current systems often require extensive subject-specific training data.
Purpose of the Study:
- To develop and evaluate user-independent intent recognition systems for powered lower limb prostheses.
- To reduce the training burden associated with current intent recognition technologies.
- To improve the accuracy and seamlessness of locomotion mode transitions.
Main Methods:
- Developed a novel mode-specific classification system treating each locomotion transition as a distinct class.
- Trained various pattern recognition algorithms using sensor data from eight lower limb amputees.
- Tested performance on a novel subject using both user-dependent and user-independent approaches.
- Incorporated sensor time history and level-ground walking data into training.
Main Results:
- Mode-specific classification reduced transitional step errors by approximately 50% for both user-dependent and user-independent systems.
- Steady-state classification accuracy was not affected by mode-specific classification.
- Including novel subject data reduced steady-state classification errors by over 60% without impacting transitional error.
- Combined strategies demonstrated significant overall system improvements compared to prior research.
Conclusions:
- User-independent, mode-specific classification is a viable strategy for powered lower limb prosthesis intent recognition.
- This approach significantly reduces errors during locomotion transitions.
- Integrating novel subject data further enhances system performance, particularly for steady-state walking.
- These advancements offer a path towards more intuitive and less burdensome prosthetic control.
