Towards real time EMG analysis

• Main Author: Kenward, Gary Wayne
• Language: English
• Published: 2010
• Online Access: http://hdl.handle.net/2429/23428

Techniques for automatic recording and analyzing single motor unit action potentials in real time have been studied. A model of the human motor unit has been derived. The muscle fibers were modelled with current dipoles and randomly distributed within the motor unit cross section according to a bivariate Gaussian, or binormal, probability density function. Computer simulations were performed with the model using parameters measured for the human biceps brachii. The single fiber action potentials generated by the current dipoles were summed at a recording site to produce a simulated single motor unit action potential. Analysis of the motor unit model and the potentials generated showed that the single motor unit action potentials of interest were recorded from well within the motor unit territory, and that the potentials formed were influenced mainly by the nearby fibers. The single motor unit action potential amplitude and duration were found to have a non-linear dependence upon the radius of the recording site from the motor unit center. The relationship between amplitude and duration was found to be linear but subject to random variations due to the influence of nearby fibers. The behaviour of the simulated single motor unit action potentials agreed with clinical observations. A three stage real time processing system was designed. The signal characteristics of clinically recorded electromyographic activity at low contraction levels were analyzed, and the first stage of pre-filtering designed to reduce the required sampling rate, and increase the signal to noise ratio without distorting the single motor unit action potentials. The second stage of processing was based on an adaptive matched filter with dual detection thresholds. The final stage used feature classification of the detected single motor unit action potentials to remove potentials not from the selected motor unit. A non-real time software implementation of the processing system demonstrated that large ensembles of single motor unit action potentials could be acquired from relatively short recordings of electromyographic activity, with few detection errors. === Applied Science, Faculty of === Electrical and Computer Engineering, Department of === Graduate