Ballistic movements: a unique approach to EMG normalizationand its effect on joint moment estimation
Date
2011
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
Modeling human movement using a computational based algorithm to predict muscle forces, ligament strains, and joint loads requires several inputs including normalized electromyographic (EMG) signals. Normalizing EMG reduces signal variability caused by electrode placement. EMG inputs scale the model’s output muscle forces and employing normalized values greater than 100% implies muscles are generating supramaximal forces. Signals from maximum voluntary isometric contractions (MVIC) are typically used for normalization, even though ballistic tasks can produce larger EMG singals. Ballistic task EMG signals normalized to MVIC EMG maximums could yield EMG values greater than 100%. Normalizing to maximal values ensures EMG signals remain below 100%. This thesis investigated dynamic signal repeatability and differences between MVIC EMG maximums and maximums from sprinting, jumping, and isokinetic movements to ascertain whether dynamic tasks produce significantly greater signals in search of a task which elicits reliable, maximal EMG. These maximum dynamic signals and the maximum MVIC values were used to normalize the model inputs. The model outputs of the two cases were compared. The results showed a significant increase of peak EMG values between the dynamic and MVIC cases for the lateral gastrocnemius (LG), medial gastrocnemius (MG), soleus (SL), vastus medialis (VM), and vastus lateralis (VL). The peak EMG was significantly greater between the ballistic and MVIC cases for the LG, MG, and SL. Divergences of signal magnitudes were repeated across testing sessions. Intra-session analysis revealed reliability in almost all peak EMG values. The two EMG-driven model outputs established the MVIC normalization procedure as the superior method. The r2 value of the knee moment for the MVIC case was 0.94 while the sprint case was 0.91 and the MVIC case produced a significant decrease in peak knee moment error, with respect to the moment calculated using inverse dynamics. This thesis identified MVIC EMG values as the better choice for normalizing dynamic EMG signals for the EMG-driven model and the results of this work establish a greater confidence in the model’s applicability to highly dynamic activities. This thesis also determined dynamic tasks exist which produce significantly larger, repeatable peak EMG signals than the MVIC tasks.