Composite Recurrent Convolutional Neural Networks Offer a Position-Aware Prosthesis Control Alternative While...

Area of Science:

• Biomedical Engineering
• Neuroscience
• Robotics

Background:

• The
• limb position effect
• hinders myoelectric prosthesis control by requiring accurate movement prediction across various limb positions.
• Current pattern recognition models rely on electromyography (EMG) and inertial measurement units, necessitating lengthy user training and frequent retraining due to signal variations.
• General-purpose models reduce training burden but offer low accuracy, while user-specific models provide high accuracy at the cost of extensive retraining.

Purpose of the Study:

• To develop an alternative control solution that mitigates the limb position effect and reduces user training requirements for myoelectric upper limb prostheses.

Main Methods:

• A novel recurrent convolutional neural network (RCNN)-based Composite Model was developed.
• This model integrates the representation learning of a general-purpose model with the decision-making of a user-specific model.
• The approach aims to leverage the strengths of both general and user-specific control strategies.

Main Results:

• The Composite Model demonstrated moderate movement predictive accuracy across diverse limb positions.
• A significant reduction in the user training routine requirements was observed compared to traditional methods.
• The findings suggest a promising new direction for prosthesis control research.

Conclusions:

• The RCNN-based Composite Model offers a balanced approach to myoelectric prosthesis control.
• It effectively addresses the challenges posed by the limb position effect and reduces the burden of model training.
• This novel approach paves the way for more intuitive and efficient prosthetic limb control.