Mitigate the Effect of Arm Posture on Electromyography Pattern Recognition.

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Signal Processing

Background:

• Electromyography (EMG)-based mechatronic devices for stroke rehabilitation face challenges with signal robustness due to arm posture variations.
• Existing methods using multiple EMG sensors require significant computational resources for real-time processing, hindering practical application.

Purpose of the Study:

• To develop a novel approach reducing the number of EMG channels for processing in rehabilitation devices.
• To improve the accuracy and speed of detecting hand grasping intention in stroke patients.

Main Methods:

• An optimal-channel-selection technique was developed, utilizing a convolutional neural network (CNN).
• The technique selects two optimal EMG channels from an eight-channel armband based on arm posture and individual demographics.
• Grasping intention prediction was performed using the selected channels and compared against an eight-channel system.

Main Results:

• The two-channel system achieved grasping intention prediction in 2.3 seconds with 81% accuracy.
• The traditional eight-channel system required 8.6 seconds for detection with 79% accuracy.
• The proposed method demonstrated faster detection and higher accuracy compared to the conventional approach.

Conclusions:

• The novel optimal-channel-selection technique effectively reduces EMG data processing requirements.
• This approach enhances accuracy and responsiveness in EMG-based mechatronic rehabilitation devices.
• The findings offer a promising solution for robust EMG signal interpretation in real-time rehabilitation settings.