Systems and methods for hierarchical pattern recognition for simultaneous control of multiple-degree of freedom movements for prosthetics

Patent metadata

Field	Value
Publication number	US-9907489-B2
Application number	US-201313918418-A
Country	US
Kind code	B2
Filing date	Jun 14, 2013
Priority date	Jun 14, 2012
Publication date	Mar 6, 2018
Grant date	Mar 6, 2018

Abstract

This disclosure relates to a hierarchy of classifiers, including linear discriminant analysis (LDA) classifiers, arranged to provide simultaneous control of multiple degrees of freedom of a prosthetic. The high accuracy of the hierarchical approach allows pattern recognition techniques to be extended to permit simultaneous control, potentially allowing amputees to produce more fluid, life-like movements, ultimately increasing their quality of life. The hierarchy may also be used to control a biological interface that allows input to computers for persons with disabilities, used as a potential video-game controller, and used as an input interface for tablets, and phones.

First claim

What is claimed is:

1. A biological interface system comprising:
a sensor comprising a plurality of electrodes for detecting a multi-cellular potential emanating from one or more living cells of a subject; and
a processing device comprising a processor and memory to extract features from the multi-cellular potential, wherein the processing device determines an intended movement of a prosthesis based on a hierarchical pattern recognition control scheme executing on the processor,
wherein the hierarchical pattern recognition control scheme comprises:
a first classifier configured to determine an intended motion of the prosthesis within a first degree of freedom of the prosthesis and to produce an output reflecting the intended motion; and
a plurality of subsequent classifiers,

wherein the hierarchical pattern recognition control scheme is configured to:
select a second classifier from the plurality of subsequent classifiers on the basis of the output of the first classifier reflecting the intended motion; and
use the second classifier to determine an intended motion of the prosthesis within a second degree of freedom of the prosthesis;

wherein the processing device is configured to send the intended movement to the prosthesis for control of the prosthesis.

2. The system of claim 1 , wherein each of the subsequent classifiers is defined by a degree of freedom of the prosthesis for a domain of possible joint movements.

3. The system of claim 2 , wherein subsequent classifiers are for a domain of possible joint movements that comprises simultaneous combinations of a set of independent movements about a single joint of the prosthesis.

4. The system of claim 2 , wherein each of the plurality of hierarchy pattern classifiers or a subset of the plurality of hierarchical pattern classifiers is conditionally dependent.

5. The system of claim 1 , wherein the multi-cellular potential is an electromyographic signal.

6. The system of claim 1 , further comprising an output unit to send the commands to a prosthetic limb for real-time control of the prosthetic limb.

7. The system of claim 1 , further comprising a controller interface, wherein the processing device further comprises an output unit to send commands to the controller interface.

8. The system of claim 7 , wherein the controller interface is selected from a group consisting of a computer, a cell phone, computer tables, and video games.

9. The system of claim 1 , wherein the hierarchical pattern recognition system has been trained using training data collected from a set of single joint movements, preprogrammed joint movements, or a combination thereof.

10. The system of claim 1 , wherein the organization of a hierarchy of the hierarchical pattern recognition system is determined by evaluating with a processing device a classification accuracy using a set of testing data.

11. The system of claim 10 , wherein an organization of the hierarchy is determined using at least one clustering algorithm.

12. The system of claim 10 , wherein at least one feature is used as an input to each classifier in the hierarchy.

13. The system of claim 12 , wherein the selection of the at least one feature is determined by evaluating with a processing device the classification accuracy using a testing set of data.

14. The system of claim 12 , wherein the selection of the at least one feature is determined by evaluating with a processing device the classification accuracy using at least one clustering algorithm.

15. The system of claim 1 , wherein the first degree of freedom of the prosthesis is a wrist flexion/extension degree of freedom.

16. The system of claim 1 , wherein the second degree of freedom of the prosthesis is a wrist pronation/supination degree of freedom.

17. The system of claim 1 , wherein the hierarchical pattern recognition control scheme is further configured to:
select a third classifier from the plurality of subsequent classifiers on the basis of an output of the second classifier that reflects the intended motion of the prosthesis within the second degree of freedom; and
use the third classifier to determine an intended motion of the prosthesis within a third degree of freedom of the prosthesis.

18. The system of claim 17 , wherein the third degree of freedom of the prosthesis is a hand open/close degree of freedom.

19. The system of claim 1 , wherein the output of the first classifier reflects an output selected from the group consisting of no intended motion within the first degree of freedom, a motion in a first direction within the first degree of freedom, and a motion in a second direction within the first degree of freedom.

20. The system of claim 1 , wherein the hierarchical pattern recognition control scheme is further configured such that if the output of the first classifier reflects no intended motion in the first degree of freedom, then the second classifier selected is one that is conditioned on a training data set that does not include combined movements involving an active motion in the first degree of freedom.

21. The system of claim 20 , wherein the hierarchical pattern recognition control scheme is further configured such that if the output of the first classifier reflects active motion in the first degree of freedom, then the second classifier selected is configured to distinguish between an active motion solely in the first degree of freedom and at least one other combined motion that includes the active motion.

22. The system of claim 1 , wherein the hierarchical pattern recognition control scheme is further configured such that if the output of the first classifier reflects active motion in the first degree of freedom, then the second classifier selected is configured to distinguish between an active motion solely in the first degree of freedom and at least one other combined motion that includes the active motion.

23. The system of claim 1 , wherein the prosthesis is an artificial prosthesis for replacement of at least a portion of a human limb.

24. The system of claim 1 , wherein the prosthesis is a virtual prosthesis displayed on a screen.

Assignees

• Chicago Rehabilitation Inst

Inventors

• Hargrove Levi
• Young Aaron

Classifications

• G16H50/70

  for mining of medical data, e.g. analysing previous cases of other patients · CPC title

• A61F2/72

  Bioelectric control, e.g. myoelectric · CPC title

• A61B5/7267 Primary

  involving training the classification device · CPC title

• A61B5/11 Primary

  Measuring movement of the entire body or parts thereof, e.g. head or hand tremor or mobility of a limb {(A61B5/1038 takes precedence; motion detection to correct for motion artifacts in physiological signals A61B5/721)} · CPC title

• A61B5/04888

  Human Necessities · mapped topic