Grasping force estimation using state-space model and Kalman filter

TL;DR

This paper introduces a novel state-space and Kalman filter-based method for estimating grasping force from surface electromyography signals, enhancing control accuracy for myoelectric prostheses.

Contribution

It develops a multivariable system identification approach combining state-space modeling and Kalman filtering for improved force estimation in prosthetic control.

Findings

Achieved $R^2$ of 0.92 indicating high correlation.

Reduced estimation error with NRMSE of 0.723.

Outperformed neural network and linear models in accuracy.

Abstract

The grip force required to handle an object depends on the object's mass and the friction coefficient of its surface. The control of grip force in myoelectric prosthesis is crucial for handling objects adequately. In the current paper we propose a new method for improving the proportional and continuous grasping force estimation to improve control systems for myoelectric prosthesis based on surface electromyography (sEMG) recordings. For this purpose, we develop an approach based on multivariable system identification in the state-space (SS) and continuous force estimation with Kalman Filter (KF). The sEMG recordings of ten healthy individuals performing a grip task were used as data set for model identification. The root mean square (RMS), the mean absolute value (MAV), and the waveform length (WL) extracted from the sEMG signals were used at the model's input while the measured grasping force was the output. The performance of the method was evaluated with the normalized root-mean-squared-error (NRMSE) and the square of the Pearson's correlation coefficient ($ R^2 $). We found the $R^2$ and NRMSE values were 0.92$ \pm $ 0.0319 and 0.723$ \pm $ 0.0563, respectively. The performance of the proposed technique was superior to the results obtained with other regression models, such as the recurrent nonlinear autoregressive exogenous (NARX)-based neural network, the multi-layer perceptron (MLP) network and the linear discriminant analysis (LDA) with a quadratic polynomial fitting (QPF). The results confirm that the method is adequate for real-time applications with myoelectric hand prostheses.