Enhancing Myoelectric Prosthetic Control: Deep Learning Strategies for Continuous Arm Kinematics Estimation and...

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Artificial Intelligence in Medicine

Background:

• Myoelectric prosthesis control has advanced but faces challenges in limited movement range and simultaneous control.
• Existing methods often struggle with precise, multi-dimensional control of prosthetic limbs.

Purpose of the Study:

• To develop an LSTM-based deep learning approach for continuous, precise control of prosthetic limb parameters.
• To enhance simultaneous control capabilities and movement range in myoelectric prostheses.
• To investigate the transferability of the control model across different users.

Main Methods:

• Utilized Long Short-Term Memory (LSTM) networks, a type of deep learning model.
• Focused on continuous control of elbow angle (θ), wrist position (X, Y), and velocity (v).
• Evaluated model performance and cross-subject transferability.

Main Results:

• Achieved enhanced precision in controlling elbow angle, wrist position, and velocity.
• Demonstrated successful continuous control of multiple prosthetic limb parameters simultaneously.
• Successfully transferred the trained model to new subjects with effective performance.

Conclusions:

• The LSTM-based deep learning approach significantly improves myoelectric prosthesis control.
• The model offers precise, simultaneous control and shows promising cross-subject transferability.
• This research advances the development of more functional and intuitive prosthetic limbs.