A CW-CNN regression model-based real-time system for virtual hand control.

Area of Science:

• Biomedical Engineering
• Rehabilitation Robotics
• Neuroprosthetics

Background:

• Myoelectric prosthetic hands are crucial for upper limb amputees' daily lives.
• Real-time control systems for prosthetics using surface electromyography (sEMG) signals remain a significant challenge.
• Existing deep learning approaches for sEMG primarily focus on offline analysis and motion classification.

Purpose of the Study:

• To develop a high-precision, real-time control system for virtual hand control using sEMG signals.
• To investigate the efficacy of a CW-CNN regression model for multi-joint angle estimation.
• To evaluate the system's performance in terms of accuracy, stability, and computational latency.

Main Methods:

• Proposed a Convolutional Wavelet Neural Network (CW-CNN) regression model for sEMG-based motion control.
• Implemented an Adaptive Kalman Filter to smooth estimated joint angles.
• Conducted real-time experiments with eight healthy participants controlling a virtual hand.

Main Results:

• The CW-CNN regression model achieved high precision in simultaneously estimating 3-Degrees of Freedom (DOFs) joint angles.
• The integrated Adaptive Kalman Filter ensured stable and smooth output of joint angle commands.
• The system exhibited low computational latency, suitable for real-time applications.

Conclusions:

• The developed real-time control system offers a promising solution for advanced myoelectric prosthetic hand control.
• The CW-CNN regression model combined with an Adaptive Kalman Filter provides accurate and efficient multi-joint control.
• Future applications may include direct integration with actual prosthetic hands for improved amputee rehabilitation.