The quantification of EMG normalization error

Abstract

Electromyography (EMG) and normalized EMG have been accepted as methods of quantifying the activity level of a muscle. Normalized EMG, in conjunction with the EMG/force relationship and muscle cross-sectional area data, allows researchers to estimate the amount of muscle force exerted across a joint. An accurate description of this muscle force is a critical input to models designed to describe the risk of injury of a task. In order to be able to make statements about the relative intensity of an EMG signal, researchers who use normalization procedures take a given EMG activity level, at a known joint angle, and relate it to some reference activity level obtained at that particular joint angle. However, there have been studies where the EMG activity of an unrestricted dynamic task, such as walking, cycling, performing an occupational task, etc., has been normalized with respect to an EMG value taken during a single maximum voluntary contraction performed at one reference joint angle. This practice will render inaccurate results because at different joint angles there are changes in the portion of the muscle within the viewing area of the electrode, as well as changes in the length/tension relationship of the muscle which would cause changes in the maximum EMG value. The present study was an attempt to quantify the errors associated with normalization relative to a reference EMG value collected at an arbitrary joint position. Four subjects performed a series of controlled trunk extension exertions. As they performed these exertions the EMG activities were collected for eight trunk muscles. The task EMG values that resulted were then: (1) all normalized with respect to the maximum EMG at a single arbitrary trunk angle and (2) each normalized with respect to that specific trunk angle's maximum EMG. The results show that for the primary trunk extensors (erector spinae) large errors (>75%) resulted from normalization using a single reference point and the magnitude of these errors followed consistent patterns within subjects as a function of trunk angle.