Towards the development of a neuroprosthetic system through which paralyzed individuals with C5 or C6 spinal chord injuries regain full functional control of their limbs, a means of estimating muscle activity level from desired kinematic values is required. This thesis presents a method of performing this estimation. It uses a priori distribution functions and Bayes' theorem to find the a posteriori distribution of possible muscle activity levels given a specified set of kinematic values. The a priori distribution functions were estimated with data taken during a training task. The system was tested by comparing predicted to actual muscle activity levels during seven different tasks, none of which was used in establishing the a priori distribution functions. The system worked well with the overall RMS error of 6.1% across all muscles and all tasks.
Chad V. Anderson, "Estimation of muscle activity using probability density functions and Bayes' theorem," MS thesis (Advisors: Andrew J. Fuglevand and Charles M. Higgins), Department of Electrical and Computer Engineering, The University of Arizona, April 2004.