On the Classification of Electromyography Signals to Control a Four Degree-Of-Freedom Prosthetic Device

Area of Science:

• Biomedical Engineering
• Rehabilitation Robotics
• Signal Processing

Background:

• Prosthetic device control quality is influenced by algorithm accuracy, control complexity, and error compensation.
• High-time-domain control granularity improves perceived error but can decrease classification accuracy.

Purpose of the Study:

• To determine the accuracy threshold for Electromyography (EMG) signal classification algorithms suitable for hand prosthetic control.
• To reduce the discrepancy between laboratory-based and commercially available prosthetic devices.

Main Methods:

• Comparison of five different EMG signal classification algorithms.
• Selection and real-time testing of the most promising algorithm.
• Performance evaluation against an Inertial Measurement Unit (IMU)-based controller.

Main Results:

• The chosen EMG-based control strategy demonstrated performance comparable to an IMU-based controller.
• The study identified a viable borderline for algorithm accuracy in prosthetic control.

Conclusions:

• EMG signal classification algorithms with efficient training times are applicable for prosthetic device control.
• EMG-based control offers a competitive alternative to IMU-based systems for hand prosthetics.