Estimating unmeasured muscle excitations using Gaussian process regression
Conference Year
January 2020
Abstract
Remote observation of patient biomechanics using wearable sensors is driving a paradigm shift in the approach to healthcare. Continuous monitoring in this way provides a pathway to personalized rehabilitation and comprehensive patient characterization. The development of practical tools for this purpose seeks to minimize patient burden. This prevents the use of traditional surface electromyography (sEMG)-based techniques for informing clinically relevant joint mechanics as they require information from all involved muscles and unwieldy sensor arrays. Recent research suggests it may be feasible to estimate unmeasured muscle excitations using sEMG data from a measured subset. This may enable the use of sEMG-based estimation of joint mechanics for the remote patient monitoring problem. However, there does not currently exist a method to perform this estimation. In this work, we present a novel Gaussian process-based model of the synergistic relationship between muscles during walking. We demonstrate the use of this approach to estimate the muscle excitation time-series of six muscles using sEMG data from four different muscles. Further, we show the superior ability of the proposed technique to reconstruct unmeasured muscle excitations compared to a non-negative matrix factorization-based approach.
Primary Faculty Mentor Name
Ryan McGinnis
Faculty/Staff Collaborators
Anna Ursiny (undergraduate collaborator), Ryan McGinnis (faculty advisor)
Status
Graduate
Student College
College of Engineering and Mathematical Sciences
Program/Major
Mechanical Engineering
Primary Research Category
Engineering & Physical Sciences
Secondary Research Category
Biological Sciences
Estimating unmeasured muscle excitations using Gaussian process regression
Remote observation of patient biomechanics using wearable sensors is driving a paradigm shift in the approach to healthcare. Continuous monitoring in this way provides a pathway to personalized rehabilitation and comprehensive patient characterization. The development of practical tools for this purpose seeks to minimize patient burden. This prevents the use of traditional surface electromyography (sEMG)-based techniques for informing clinically relevant joint mechanics as they require information from all involved muscles and unwieldy sensor arrays. Recent research suggests it may be feasible to estimate unmeasured muscle excitations using sEMG data from a measured subset. This may enable the use of sEMG-based estimation of joint mechanics for the remote patient monitoring problem. However, there does not currently exist a method to perform this estimation. In this work, we present a novel Gaussian process-based model of the synergistic relationship between muscles during walking. We demonstrate the use of this approach to estimate the muscle excitation time-series of six muscles using sEMG data from four different muscles. Further, we show the superior ability of the proposed technique to reconstruct unmeasured muscle excitations compared to a non-negative matrix factorization-based approach.