Using Hidden Markov Models to track upper extremity arm motions for surface electromyographic based robot teleoperation

Abstract

A Hidden Markov Model (HMM) is used to predict and characterize a stochastic process that is not easily identifiable; i.e., it is hidden from the observer. This process can only be identified through an additional set of stochastic events that is not only observable, but is also responsible for producing the original hidden stochastic process mentioned above. The goal of this project is to use a HMM to track upper-extremity arm motions performed in the sagittal plane (representing the hidden states) by means of the surface electromyographic (sEMG) activity associated with these arm motions (representing the observed states). After which, we intend to use the characterized sEMG signals to teleoperate a robotic manipulator. We wish to create a rehabilitative robotic platform for people who have suffered from progressive muscular degenerative disorders and neurological deficits. This platform will take advantage of any residual physiological information that is still available (non-invasively) within these individuals. The ultimate goal is to create more intelligent orthotics and wearable robotic systems for people having these types of disabilities. It is hoped that this kind of device could assist a disabled user in performing daily living tasks that require reaching for an object. The HMM algorithm presented here is implemented and tested offline in Matlab with five healthy participants. It was successful in tracking two degrees of freedom on the human arm (representing the elbow and shoulder joints) with less than 15° of error.