Enhanced Performance for Multi-Forearm Movement Decoding Using Hybrid IMU-sEMG Interface.

Area of Science:

• Biomedical Engineering
• Rehabilitation Robotics
• Human-Computer Interaction

Background:

• Surface electromyography (sEMG) offers a promising avenue for intuitive prosthetic hand control.
• Current commercial prosthetic hands lack intuitive control due to variations in sEMG signals in dynamic environments.
• Bridging the gap between academic research and industry requires addressing real-world signal variability.

Purpose of the Study:

• To investigate the impact of arm position variation on sEMG signal classification for hand motion.
• To compare static versus dynamic movement strategies for training hand motion classifiers.
• To evaluate the effectiveness of sensor fusion techniques combining sEMG and inertial measurement units (IMUs).

Main Methods:

• A wearable system captured sEMG signals and forearm position using IMUs during six hand motions.
• Features were extracted from sEMG signals in the time domain (TD) and combined with position-aware (POS) data.
• Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM) classifiers were trained and tested using TD and TD-POS features under static and dynamic conditions.

Main Results:

• SVM classification performance showed a significant difference between static and dynamic training approaches.
• TD-POS features demonstrated enhanced classification accuracy compared to TD features alone.
• Dynamic training strategies and sensor fusion improved prosthetic control system performance.

Conclusions:

• Dynamic training approaches are more effective for real-world prosthetic hand control.
• Integrating arm position information (sensor fusion) significantly boosts classification performance.
• These findings pave the way for more intuitive and effective sEMG-based prosthetic control systems.