Related Experiment Video
Updated: Jun 19, 2026
11:16
Engineering Platform and Experimental Protocol for Design and Evaluation of a Neurally-controlled Powered Transfemoral Prosthesis
Published on: July 22, 2014
Multiclass real-time intent recognition of a powered lower limb prosthesis
Huseyin Atakan Varol1, Frank Sup, Michael Goldfarb
1Department of Mechanical Engineering, Vanderbilt University, Nashville, TN 37235, USA. atakan.varol@vanderbilt.edu
IEEE Transactions on Bio-Medical Engineering
|October 23, 2009
Summary
This study presents a novel control system for prosthetic legs that recognizes user intent (walking, standing, sitting) using only prosthesis sensors. This approach enables seamless transitions without needing sensors on the intact leg.
Area of Science:
- Biomedical Engineering
- Robotics
- Rehabilitation Engineering
Background:
- Powered lower limb prostheses require intuitive control for user mobility.
- Existing control systems often rely on complex sensor arrays or invasive instrumentation.
- Real-time intent recognition is crucial for seamless prosthetic function.
Purpose of the Study:
- To develop and validate a real-time intent recognition system for powered lower limb prostheses.
- To enable supervisory control based on user intent without sound-side leg instrumentation.
- To improve the naturalness and usability of prosthetic limb control.
Main Methods:
- A control architecture and intent recognition approach using prosthesis sensor data.
- Feature extraction from prosthesis signals, followed by dimensionality reduction.
- Training intent models to classify user activities (standing, sitting, walking) in real time.
Main Results:
- The system accurately identified 90 out of 90 intended activity-mode transitions in a unilateral transfemoral amputee.
- Transitions were switched without perceivable delay to the user.
- Six unintended transitions were misclassified but did not negatively impact functionality or user perception.
Conclusions:
- The developed control architecture and intent recognition approach are effective for real-time supervisory control of powered lower limb prostheses.
- The system demonstrates the feasibility of intent recognition using only prosthesis-mounted sensors.
- This technology holds promise for enhancing the mobility and user experience of amputees.
