Comparison of six electromyography acquisition setups on hand movement classification tasks
Stefano Pizzolato1,2, Luca Tagliapietra2, Matteo Cognolato1,3
1Information Systems Institute at the University of Applied Sciences Western Switzerland (HES-SO Valais), Sierre, Switzerland.
Plos One
|October 13, 2017
Summary
Surface electromyography (sEMG) hand prostheses show promise but face challenges. This study compares six sEMG acquisition setups, finding comparable results for three, including an affordable option, aiding researcher choices.
Area of Science:
- Biomedical Engineering
- Rehabilitation Technology
- Human-Computer Interaction
Background:
- Surface electromyography (sEMG) offers a non-invasive method for controlling hand prostheses, enhanced by machine learning.
- Dexterous prostheses remain underutilized due to control complexities, reliability issues, and high costs.
- A variety of sEMG acquisition systems exist, with prices varying significantly.
Purpose of the Study:
- To conduct a comparative analysis of six different sEMG acquisition setups.
- To evaluate their performance on a standardized hand movement classification task.
- To provide guidance for researchers in selecting appropriate sEMG acquisition systems based on cost and performance.
Main Methods:
- Six sEMG acquisition setups were evaluated, utilizing four distinct electrode types (Otto Bock, Delsys Trigno, Cometa Wave + Dormo ECG, and two Thalmic Myo armbands).
- Standardized protocols were employed to record over 50 distinct hand movements from healthy subjects.
- A consistent feature extraction and data analysis pipeline was applied for hand movement classification across all setups.
Main Results:
- Comparable classification performance was achieved by three sEMG acquisition setups: Delsys Trigno, Cometa Wave, and an economical dual Myo armband configuration.
- The findings indicate that effective sEMG-based research is feasible even with budget constraints, such as in smaller labs or for pediatric applications.
- Publicly available datasets allow for easy comparison and offline testing of the evaluated systems.
Conclusions:
- The study demonstrates that cost-effective sEMG acquisition systems can yield performance comparable to more expensive options for hand movement classification.
- This research facilitates informed decisions for researchers and developers working on sEMG-controlled prosthetic devices.
- The availability of standardized datasets promotes further research and development in the field of prosthetic control.
