Related Experiment Video
Updated: Jun 28, 2025
09:41
Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping
Published on: April 21, 2023
1.6K
sEMG-Based Robust Recognition of Grasping Postures with a Machine Learning Approach for Low-Cost Hand Control.
Marta C Mora1, José V García-Ortiz1, Joaquín Cerdá-Boluda2
1Department of Mechanical Engineering and Construction, Universitat Jaume I, Avda de Vicent Sos Baynat s/n, 12071 Castelló de la Plana, Spain.
Sensors (Basel, Switzerland)
|April 13, 2024
Summary
This study introduces a machine learning (ML) approach for real-time hand posture recognition using surface electromyography (sEMG) signals. This method enables affordable functional prostheses and social robots by interpreting human intentions.
Area of Science:
- Biomedical Engineering
- Robotics
- Machine Learning
Background:
- Artificial hand design faces challenges due to complex control and high costs of advanced prostheses.
- Effective human-robot interaction (HRI) for social robots requires accurate interpretation of human intention.
- Machine learning (ML) is underutilized for grasping posture recognition, limiting prosthetic and robotic capabilities.
Purpose of the Study:
- To develop and validate an ML-based system for real-time recognition of nine essential hand postures.
- To enable affordable, functional artificial hands and improve HRI for social robots.
- To achieve robust and efficient grasping posture recognition with minimal computational resources.
Main Methods:
- Utilized surface electromyography (sEMG) data from 20 subjects using a Myo armband (NinaPro DS5 and YCB Object Set).
- Developed a simple multi-layer perceptron model in MATLAB, optimized using GPU-based implementations.
- Focused on feature selection, generalization, robustness to electrode shift, and real-time performance.
Main Results:
- Achieved a global success rate of 73% in recognizing nine hand postures with only two features.
- Identified an optimal ML architecture suitable for low-cost device implementation.
- Demonstrated robustness and efficiency for real-time applications.
Conclusions:
- The proposed ML approach offers a viable solution for low-cost grasping posture recognition.
- This technology can significantly advance the development of affordable functional prostheses.
- Enhances the potential for seamless human-robot interaction in social robotics.
