Estimation of Knee Extension Force Using Mechanomyography Signals Detected Through Clothing

This paper proposes a more flexible method for estimating the knee extension forces using mechanomyography (MMG) signal. We detect the MMG signal of the quadriceps through clothing. Then several features were extracted from three channels. A support vector machine model was applied to generate mapping between the features and the actual forces. Results indicates that the best estimation performance can be obtained using a combination of four features (the mean absolute value (MAV), the sample entropy (SampEn), the mean power frequency (MPF) and the correlation coefficients of 2 different channels (CC2Cs)). Finally, an average coefficient of determination (R2) of 0.799 and a root-mean-squared error (RMSE) of 9.18% of the maximum voluntary isometric contraction (MVC) were obtained. These results are similar to those of earlier studies, which suggests that the information in the MMG signal has not been detectably reduced when measured through clothing. Therefore, the MMG signal through clothing is able to function as a reliable estimator of muscle contraction strength and can be used as the muscle machine interface to control robotics or to monitor human activity.