Shape analysis and clustering of Surface EMG Data

Abstract

Functional Data Analysis (FDA) is a recent field in data analysis and processing. It provides efficient methods and tools by considering the analyzed data as realizations of functions. In this discipline, raised shape analysis approaches. Among them, the Core Shape Modelling (CSM) furnished statistical tools for the evaluation of the shape dispersion among a set of curves. In this work, it is proposed to use this approach to study Surface EMG (SEMG) Data. These data represent electrical activity elicited during muscle contractions and measured on the surface of the skin. The generation of the SEMG signal is dependent on many morphological, physiological and neural parameters. In fact, the neural parameters tune the spatial and time recruitment of the Motor Units (MUs). In this study, the CSM algorithm is applied to detect MUs firing synchrony on SEMG data simulated using a realistic generation model. The generation parameters induce several variabilities and compensatory effects on SEMG data that could complicate and bias the data processing task. After phase realignment, a shape clustering is done on SEMG amplitude histograms using CSM formalism for different MU synchrony classes. The obtained results are promising and demonstrate the ability of shape analysis using the CSM approach to detect and classify MUs firing synchrony levels in SEMG data despite the present variabilities.