EMG Signal Decomposition Methods for Physical Actions: A Comparative Analysis

Advanced approaches are needed for the detection, decomposition, and classification of EMG data obtained from muscles. This study evaluates three popular decomposition techniques, Discrete Wavelet Transform (DWT), Empirical Mode Decomposition (EMD), and Variational Mode Decomposition (VMD), in order to determine which one offers superior accuracy. This paper uses the dataset of physical actions (Handshaking, Clapping, and Hammering). First, decomposition is done using DWT, EMD, and VMD. Then six time-domain features (Standard Deviation, Mean Absolute Value, Kurtosis, Root Mean Square Value, Waveform Length, and Willison Amplitude) are extracted. In the last stage, classification accuracy is calculated using a linear discriminant (LDA) classifier for feature vectors using principal component analysis (PCA). Three decomposition techniques for EMG signals are compared using the accuracy results.