Research Areas   >   Neuromuscular Control and Plasticity Lab    |    Research    |    Publications    |    People    |    Collaborators

Neuromuscular Control and Plasticity Lab: Selecting Optimal Inputs for Neural Machine Interfaces

This research addresses several issues that arise from the development of neural-machine interfaces. The term neural-machine interface (NMI) refers to devices that link physical or neurological signals to patient intent for movement or communication augmentation. NMIs may need information extraction from multiple signal sources (e.g. electromyogram [EMG] electrodes) and may detect intent for multidimensional control (e.g. 3D reaching). This leads to multiple-input, multiple-output (MIMO) systems. MIMO systems are computationally complex, and large numbers of inputs results in large computational burden. By reducing input number, we can improve computational/communication efficiency, algorithm performance, and practicality of use.

This study has two main goals: to extend previous input selection algorithms to systems with multiple outputs, so that they can be applied for interpretation of multiple movements; and to evaluate how the algorithm performs on systems lacking explicit output, so they can be applied in disabled patients who can only ‘imagine’ movements instead of physically performing them. Experimental methods for testing algorithm performance on a MIMO system include recording EMGs and joint angle motion data while healthy subjects perform simple arm movements guided by a visual cueing system (Figure A). To replicate situations where patients lack explicit output, the selection algorithm is run on systems consisting of either the subject’s own EMGs and joint angles, or the subject’s own EMGs and a different subject’s joint angles. We found that predicting joint angles using optimal inputs selected by the algorithm resulted in better performance than predictions using randomly selected inputs, regardless of selecting with subjects’ own angles or with alternate subject angles (Figure B).

 


home    |    people    |    research    |    contact us    |    links

Copyright © 2014 The Sensory Motor Performance Program. All Rights Reserved. Site designed by Academic Web Pages.