Principal Component Analysis of Electromyographic Signals: An Overview

Abstract

Surface electromyography (EMG) is a widely used, straight-forward, technique which allows to investigate patterns of neuromuscular activation. In contrast to the relative simplicity of the recording technique, the analysis of the derived electric signals may be rather sophisticated. The last decade, in particular, has been characterized by the development of a several quantitative approaches to the analysis of the EMG signals. The common principle underlying these analyses is the decomposition of the EMG signal waveforms in a small set of basis waveforms that capture most of the relevant features of the source EMGs and define a low-dimensional space on which the original EMG activation patterns can be represented as vectors. This could be particularly useful when the aim is to classify quantitatively EMG patterns recorded across muscles or from the same muscle across several motor tasks. Within this framework, this article will be focused on one of these approaches, the Principal Component Analysis, which has a strong potential for large scale diffusion both in research and clinical settings because of its conceptual simplicity and high practicality. The intent is to provide an overview/tutorial of the PCA applied to surface EMG signals, first by outlining the main methodological aspects and, then, by drawing examples from the movement control literature where PCA has been used effectively to gain insight on the neural processes that may underlie the control of common actions of our motor repertoire such as arm pointing and gait.