Classification of upper arm EMG signals during object-specific grasp

Area of Science:

• Biomedical Engineering
• Neuroscience
• Robotics

Background:

• Electromyographic (EMG) signals offer a viable method for controlling prosthetic limbs.
• Traditional EMG control relies on complex pattern recognition due to muscle loss after amputation, leading to less natural control.
• Recent research suggests muscle activation timing during reaching movements may predict grasping tasks.

Purpose of the Study:

• To investigate if muscle activation onset/offset during reaching can predict grip types in humans.
• To explore a more natural EMG-based control strategy for prosthetic devices.
• To validate findings from animal studies in human subjects.

Main Methods:

• Developed a support vector machine (SVM) based pattern recognition algorithm.
• Recorded EMG signals from proximal and distal muscles during reach-to-grasp movements.
• Analyzed muscle activation onset and offset patterns in three able-bodied subjects.

Main Results:

• The SVM algorithm successfully predicted grip types based on EMG timing information.
• Muscle activation onset/offset patterns during the reaching phase contained discriminative information for different grasp types.
• Preliminary results indicate the feasibility of this approach in human subjects.

Conclusions:

• Muscle activation timing during reaching movements is a promising feature for natural EMG-based prosthetic control.
• This approach could lead to more intuitive and acceptable prosthetic hand control strategies.
• Further research is warranted to confirm these findings and develop clinical applications.