Abstract:The surface electromyogram (EMG) is used to infer the neural drive sent by the central nervous system to the muscles by associating its amplitude to the strength of the neural activation. In addition, the surface EMG is adopted in the design of technologies for motor rehabilitation with the aim to restore the lost limb functionality of amputees. In this perspective, it is possible to translate this biologic signal into commands for controlling intuitively the prosthetic device. Since the surface EMG amplitude constitutes a poor indicator of the neural information, more advanced methods were developed such as the high-density surface EMG signal decomposition. With this work, I focused my efforts on understanding how the central nervous system (CNS) controls the muscle force, by exploiting the discharge timings of the motor units. This might help to lay the foundations for an innovative human-machine interface directly based on the spinal cord circuitry information. In this context, the aim of this thesis is to compare the use of the traditional surface EMG signal versus the offline high-density surface electromyography (HDsEMG) signal decomposition. For this study, I used a Graphical User Interface (GUI )which implement Convolutive Kernel Convolution (CKC). This graphical tool allows the operator to decompose the EMG signal and to perform a manual correction of the possible algorithm errors. Thanks to the decomposition approach, the neural activation sent by the spinal cord to the muscles was explored. The entire analysis is focus on the activity of the tibialis anterior muscle (TA) which was recorded by using a grid of 64 electrodes (HDsEMG). In each experiment, the participants performed 25 isometric ankle dorsiflexion at three different levels of speed (i.e.%MVC/s). In particular, the study consists of two main part: 1. the investigation of the motor units recruitment and the discharge rates modulations to understand how the CNS responds to the different conditions, such as the changing in the contraction speed; 2. the design of two different linear regressor, based on the neural drive and the EMG envelope respectivelly, to estimate the force of the contraction. In particular three single linear regressor (SLR) and the global linear regressor (GLR) were used. Both the regressions were done considering a time domain feature extracted from the recording force, i.e. the contractile impulse (IC). This was calculated as the integral of the force along time. Thus for both the EMG envelope and the neural drive, the regressions were evaluated with the integral of the signals in each contraction, instead of the signal itself. In order to investigate the performance of the regressors, the R2 value was evaluated and statistically compared among the different cases. A higher R2 mean values was obtained with the single linear ND model based regressor and the global ND model-bases regressor. Results suggest that using the HDEMG decomposition similar or even better performance can be achieved for estimating the contractile force of a muscle. Limitsof this work, due to time and facilities constraints, are mainly related to the numerosity of subjects involved for the analysis. Future development should also investigate dynamic contractions and different muscles.

Machine Learning for Detection of Muscular Activity from Surface EMG Signals

Comparison of machine learning algorithms and feature extraction techniques for the automatic detection of surface EMG activation timing

Automatic detection of surface EMG activation timing using a wavelet transform based method

Denoising Surface Electromyography Signals Based on Complementary Ensemble Empirical-Mode Decomposition and Multi-Scale Entropy

Prediction of Lower Limb Action Intention Based on Surface EMG Signal

MuscleNET: mapping electromyography to kinematic and dynamic biomechanical variables by machine learning

A novel technique for muscle onset detection using surface EMG signals without removal of ECG artifacts

Real-Time Forecasting of sEMG Features for Trunk Muscle Fatigue Using Machine Learning

Detection of movement onset using EMG signals for upper-limb exoskeletons in reaching tasks

Muscle fatigue detection based on sEMG signal using autocorrelation function and neural networks

A pilot study on the daily control capability of s-EMG prosthetic hands by amputees

Embedded Machine Learning System for Muscle Patterns Detection in a Patient with Shoulder Disarticulation

High density EMG signal decomposition for force prediction: perspective for man-machine interfacing.

Continuous reach-to-grasp motion recognition based on an extreme learning machine algorithm using sEMG signals

Innovative use of optical sensors for real-time imaging and diagnosis using enhanced verge denoising and LSTM for sports medicine

Recognition of Muscular Activity in Complex Noise-Contaminated Myoelectric Signals: An Adaptive Two-Step Approach

Can Data-Driven Supervised Machine Learning Approaches Applied to Infrared Thermal Imaging Data Estimate Muscular Activity and Fatigue?

Machine learning allows expert level classification of intraoperative motor evoked potentials during neurosurgical procedures

AML-DECODER: Advanced Machine Learning for HD-sEMG Signal Classification—Decoding Lateral Epicondylitis in Forearm Muscles

Comparative Analysis of Supervised Classifiers for the Evaluation of Sarcopenia Using a sEMG-Based Platform

Distinguishing normal, neuropathic and myopathic EMG with an automated machine learning approach