Statistical analysis of EMG-based features for different hand movements

Abstract

The electrical activity of the muscles is analyzed by surface Electromyography (sEMG). EMG signals are the essential source of control for upper limb prosthetics and orthotics and also find numerous applications in biomedical engineering and rehabilitation fields. This work focuses on the analysis of sEMG signals acquired for three different hand actions using Analysis of Variance (ANOVA) for understanding the variability of features. A single-channel sEMG amplifier is designed and signals are recorded for three different hand movements from normal subjects. Empirical Mode Decomposition (EMD) is applied to denoise the signal from artifacts. Features are extracted in time, spectral, and wavelet domain. The prominent features are selected using fuzzy entropy measure. ANOVA on prominent features shows a linear relationship between features and different hand movements and therefore these prominent features can be used to activate the prosthetic hand.