Enhanced EMG signal processing for simultaneous and proportional myoelectric control

Area of Science:

• Biomedical Engineering
• Neuroscience
• Robotics

Background:

• Prosthetic limb control remains a challenge, particularly for multi-degree-of-freedom (DOF) devices.
• Surface electromyographic (sEMG) signals offer a promising, non-invasive source for intuitive prosthetic control.
• Existing sEMG processing methods often struggle with the complexity of multi-DOF control.

Purpose of the Study:

• To develop and evaluate a new signal processing scheme for extracting neural control information from multi-channel sEMG.
• To enable proportional control of multi-DOF prosthetic devices.
• To improve the accuracy and robustness of myoelectric control systems.

Main Methods:

• Extracted four time-domain (TD) features from multi-channel sEMG during isometric wrist contractions across three DOFs.
• Collected corresponding force data using a custom sensor.
• Trained a multilayer perceptron (MLP) neural network using extracted features and force signals with five-fold cross-validation.

Main Results:

• The proposed signal processing scheme successfully extracted neural control information from sEMG signals.
• The MLP model demonstrated effective learning of the relationship between sEMG features and wrist forces.
• The new method showed significant performance improvement compared to a previous sEMG processing approach for multi-DOF control.

Conclusions:

• The developed signal processing scheme is effective for extracting neural control information from sEMG for multi-DOF prosthesis control.
• This approach offers a significant advancement in proportional myoelectric control.
• The findings pave the way for more intuitive and functional prosthetic limb systems.