Implementation of a Real-Time Force Estimation System Based on sEMG Signals and Gaussian Process Regression: Human–Robot Interaction in Rehabilitation

Implementation of a Real-Time Force Estimation System Based on sEMG Signals and Gaussian Process Regression: Human–Robot Interaction in Rehabilitation

Human force estimation has numerous applications, including biomedical models, rehabilitation, biomechanical system control, and human-machine interfaces. To enable such applications, it is necessary and challenging to develop methods for efficiently estimating force. In this work, we propose a syst...

Full description

Saved in:
Bibliographic Details
Main Authors: Thantip Sittiruk, Kiattisak Sengchuai, Apidet Booranawong, Pornchai Phukpattaranont
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
System implementation
electromyography
force estimation
regression analysis
patient rehabilitation
Online Access:https://ieeexplore.ieee.org/document/10843210/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Human force estimation has numerous applications, including biomedical models, rehabilitation, biomechanical system control, and human-machine interfaces. To enable such applications, it is necessary and challenging to develop methods for efficiently estimating force. In this work, we propose a system for muscle force estimation using eight-channel surface electromyography (sEMG) signals. The main contribution of this work is the implementation of a real-time force estimation system in MATLAB/Simulink to support real-time testing and usability. The regression models (i.e., Gaussian process regression (GPR), neural networks (NN), linear regression (LR), and support vector machines (SVM)) are applied to estimate the forearm muscle forces exerted by different elbow placement patterns for rehabilitation applications. The GPR model with exponential kernel function is chosen as the best model and is used to calculate forces at various elbow placement scenarios. Experimental results show that the GPR model can estimate forces on the X and Y axes concurrently in four movement directions: X+, X-, Y+, and Y-. The average root mean square errors (RMSEs) indicating estimation performance for offline and online testing of different elbow placement patterns are 1.09 to 1.99 N and 3.58 to 4.16 N, respectively.
ISSN:2169-3536

Similar Items