Application of Principal Components Analysis for Evaluation and Classification of Complex Emg Data

Many biomechanical models used to produce injury risk estimates for the lower trunk require lower trunk muscle forces as inputs. These forces are typically estimated through the use of surface electromyography (sEMG). The variability inherent in sEMG measurements can, and should, be analyzed to determine the possible presence and sources of excessive variation in the data. Principal components analysis (PCA) provides a robust and straightforward method for performing an analysis of the variability of complex sEMG datasets. This paper describes the results obtained from the application of PCA to a dataset consisting of activation levels for several lower trunk muscles. The results demonstrate the value of the technique in identifying clusters of observations in the data and in simplifying the multidimensional dataset. The use of PCA as a hypothesis generation tool is also explored.