Motion recognition for simultaneous control of multifunctional transradial prostheses.

Area of Science:

• Biomedical Engineering
• Rehabilitation Robotics
• Human-Computer Interaction

Background:

• Myoelectric prostheses rely on electromyography (EMG) pattern recognition for control.
• Current systems often struggle with classifying combined upper-limb movements due to single-movement focus.
• This limits the functionality of prostheses in real-world daily activities.

Purpose of the Study:

• To develop and evaluate improved EMG pattern-recognition classifier training schemes.
• To address the limitations of current methods in identifying combined upper-limb motions.
• To enhance the performance of multifunctional myoelectric prosthesis systems.

Main Methods:

• Proposed four novel classifier training schemes for EMG pattern recognition.
• Investigated the effectiveness of these schemes in identifying combined upper-limb motions.
• Evaluated classification performance in both able-bodied subjects and transradial amputees.

Main Results:

• Three of the four proposed training schemes significantly improved classification performance.
• Achieved average classification accuracies of 75.10%-77.56% in able-bodied subjects.
• Achieved average classification accuracies of 62.50%-63.38% in transradial amputees.

Conclusions:

• The proposed training schemes offer superior classification performance for combined motions compared to existing methods.
• These advancements hold promise for more intuitive and functional myoelectric prosthesis control.
• Improved classification accuracy can lead to enhanced user experience and independence for amputees.