Application of neural based estimation algorithm for gait phases of above knee prosthesis

Area of Science:

• Biomedical Engineering
• Robotics
• Prosthetics

Background:

• Semi-active knee prostheses require advanced control systems to adapt to user demands and environmental conditions.
• Accurate gait phase estimation is crucial for responsive prosthetic function.

Purpose of the Study:

• To develop and compare two gait phase estimation methods for a microcontroller-based semi-active knee prosthesis.
• To evaluate the performance of rule-based quantization versus an artificial neural network (ANN) model.

Main Methods:

• Developed and implemented rule-based quantization and ANN-based gait phase estimation algorithms.
• Collected synchronous gait data using inertial measurement systems (accelerometers, gyroscopes) and image-based systems.
• Utilized a microcontroller for real-time data processing and prosthesis control.

Main Results:

• Both methods dynamically normalized input data for gait phase estimation.
• The embedded ANN-based approach demonstrated slightly superior performance compared to the rule-based algorithm.
• The ANN approach offers better scalability for incorporating additional input parameters within microcontroller constraints.

Conclusions:

• The ANN-based gait phase estimation method is a promising approach for enhancing semi-active knee prosthesis adaptability.
• The findings suggest that ANN models can be effectively implemented on microcontrollers for real-time prosthetic applications.