Research on Upper Limb Motion Intention Classification and Rehabilitation Robot Control Based on sEMG

Area of Science:

• Biomedical Engineering
• Neuroscience
• Robotics

Background:

• Surface electromyography (sEMG) is a non-invasive technique measuring muscle electrical activity.
• sEMG reflects motor intentions and muscle contraction levels, crucial for human-machine interfaces.
• Upper limb movement control is vital for rehabilitation robotics.

Purpose of the Study:

• To classify and recognize nine types of upper limb motor intentions using sEMG.
• To apply these recognized intentions for interactive control of an end-effector rehabilitation robot.
• To compare the efficacy of traditional and deep learning methods for sEMG classification.

Main Methods:

• Data acquisition and preprocessing of sEMG signals from upper limb muscles.
• Development and validation of an upper limb musculoskeletal model in OpenSim.
• Implementation and comparison of traditional machine learning and deep learning models (MLCNN) for nine-class sEMG intention recognition.
• Integration of MLCNN-based intention recognition for controlling the iReMo® rehabilitation robot.

Main Results:

• Machine learning and deep learning methods achieved high classification accuracy for nine upper limb motor intentions.
• The multi-stream convolutional neural network (MLCNN) demonstrated strong performance in extracting motor intentions.
• The MLCNN-controlled robot system exhibited smooth and accurate motion control across various trajectories.

Conclusions:

• sEMG-based motor intention recognition is effective for controlling rehabilitation robots.
• Deep learning approaches, particularly MLCNN, offer a robust method for decoding complex motor intentions.
• This technology holds promise for enhancing interactive and intuitive control in upper limb robotic rehabilitation.