Related Experiment Video
Updated: May 16, 2026
08:15
Capturing Dynamic Finger Gesturing with High-resolution Surface Electromyography and Computer Vision
Published on: March 28, 2025
SEMG-based hand motion recognition using cumulative residual entropy and extreme learning machine.
1School of Communication and Information Engineering, Shanghai University, Shanghai, China. junshi@staff.shu.edu.cn
Medical & Biological Engineering & Computing
|December 11, 2012
Summary
This study introduces a new method using cumulative residual entropy (CREn) and extreme learning machines (ELM) for hand motion recognition from surface electromyography (SEMG) signals. The CREn-ELM approach offers high accuracy and computational efficiency for potential use in prosthetic control.
Area of Science:
- Biomedical Engineering
- Signal Processing
- Machine Learning
Background:
- Surface electromyography (SEMG) signals are crucial for understanding and controlling prosthetic devices.
- Accurate recognition of multiple hand motions from SEMG is challenging due to signal complexity and noise.
- Existing feature extraction and classification methods have limitations in speed and accuracy.
Purpose of the Study:
- To propose and evaluate a novel scheme for recognizing multiple hand motions using SEMG signals.
- To introduce cumulative residual entropy (CREn) as a feature extraction method.
- To employ the extreme learning machine (ELM) classifier for enhanced motion discrimination.
Main Methods:
- Utilized cumulative residual entropy (CREn) for feature extraction from SEMG signals.
- Employed extreme learning machine (ELM), a fast feedforward neural network, for classification.
- Compared CREn against fuzzy entropy, sample entropy, approximate entropy, and time-domain features.
- Benchmarked ELM against linear discriminant analysis (LDA) and support vector machine (SVM).
- Tested the system on four-channel SEMG data from ten subjects.
Main Results:
- CREn achieved superior classification accuracy compared to other entropy measures across all classifiers.
- CREn performance was comparable to state-of-the-art time-domain features for various segment lengths.
- CREn exhibited lower computational complexity than alternative features.
- ELM demonstrated significantly faster processing speeds than LDA and SVM without compromising accuracy.
- The proposed CREn-ELM scheme showed high potential for real-time applications.
Conclusions:
- The CREn-ELM scheme provides an effective and efficient method for multi-hand motion recognition from SEMG.
- CREn is a robust feature for capturing signal uncertainty in SEMG analysis.
- ELM offers a computationally advantageous classification alternative for SEMG-based systems.
- This approach holds promise for advancing the control of SEMG-based multifunctional prostheses.
