Metodologia para detecção automática da ativação muscular em sinais eletromiográficos

Abstract

This study proposes a new methodology for automatic detection of muscle activation in electromyographic (EMG) signals. Which uses the local variance of EMG signal to determine the onset and offset times of muscle activation events. Were implemented two existing and consolidated methods (Teager-Kaiser Energy Operator and Sample Entropy) in order to carry the activation detection by another way and enable a comparative analysis of different methodologies. The evaluation of results was separated into two stages: performance analysis and convergence analysis. The performance analysis was established by quantifiable and objective parameters: accuracy, tolerance to noise and computational cost. It was developed also a generator of synthetic EMG signals whose muscle activation times and signal to noise ratio (SNR) were previously known. Considering the parameters established and the data analyzed, the proposed methodology demonstrated a better precision and tolerance to noise when compared to the others methods. The convergence analysis used the real EMG data from ten subjects, of which was collected signals from eight different muscles. Through this set of data, it was possible to demonstrate the high correlation between the results from the analyzed methods.