Comparison of surface and intramuscular EMG pattern recognition for simultaneous wrist/hand motion classification

Area of Science:

• Biomedical Engineering
• Neuroprosthetics
• Rehabilitation Robotics

Background:

• Intuitive control of artificial limbs requires simultaneous management of multiple degrees of freedom (DOFs).
• Electromyogram (EMG) signals are crucial for decoding motor intentions for prosthetic limb control.
• Comparing surface EMG and intramuscular EMG is essential for optimizing prosthetic control systems.

Purpose of the Study:

• To evaluate if intramuscular EMG enhances pattern classification accuracy for simultaneous wrist/hand movements compared to surface EMG.
• To assess the performance of two distinct pattern classification methods using different EMG signal sources.
• To determine the impact of intramuscular EMG on classifying both 1-DOF and 2-DOF movements.

Main Methods:

• Utilized two pattern classification methods: a single classifier for 1-DOF/2-DOF discrimination and a parallel set of three classifiers for individual DOF prediction.
• Trained classifiers to predict movements including wrist rotation, wrist flexion/extension, and hand open/close.
• Compared classification errors between intramuscular EMG and surface EMG for both classification configurations.

Main Results:

• Intramuscular EMG significantly reduced classification error compared to surface EMG in the parallel classifier configuration (p<0.01).
• No significant difference in classification error was observed between intramuscular and surface EMG for the single classifier method.
• Intramuscular EMG mitigated error increases when the parallel classifier was trained without 2-DOF motion data.

Conclusions:

• Intramuscular EMG offers superior performance for complex, multi-DOF prosthetic hand and wrist control, particularly with parallel classification strategies.
• The findings suggest intramuscular EMG is a more robust signal for advanced prosthetic control systems requiring fine-tuned movement decoding.
• This research advances the development of more intuitive and life-like artificial limb control through improved EMG signal processing.