Soft Sensor for Hand-Grasping Force by Regression of an sEMG Signal

This paper presents the implementation of a soft sensor for hand-grasping force by the sEMG (Surface Electromyography) collect from 6 different muscles in the ventral regions of the forearm. This work is implemented in the envelope of the signal from sEMG by a low-pass filter with a cut frequency of \(3 {\text {Hz}}\), which maintains the information of the energy of the signal. An Artificial Neural Network (ANN) was applied for the regression of the force and was done an online application of the model as a soft sensor, and has as input the 6 channels of sEMG rectified and filtered. Four volunteers were tested to see the viability of the regression, all of them showed high \(R_{{\text {sq}}}\) for fitting the regression model, 0.99, 0.98, 0.98 and 0.97, respectively proving the capability of the application. The online performance demonstrated \(16.66 [{\text {N}}]\) of root mean square error, approximately \(3.14\%\) of a MVC (Maximal Voluntary Contraction) threshold of volunteer 01.