System and method for multifunctional control of upper limb prosthesis via EMg signal identification

• US 4,209,860 A
• Filed: 02/13/1978
• Issued: 07/01/1980
• Est. Priority Date: 02/13/1978
• Status: Expired due to Term

• Current Assignee: The United States of America As Represented By The Secretary of Agriculture
• Original Assignee: The United States of America As Represented By The Secretary of Agriculture
• Inventors: Graupe, Daniel
• Primary Examiner(s): Frinks, Ronald L.
• Application Number:
  US05/877,379
• Time in Patent Office: 869 Days
• Field of Search: 3/1.1, 3/12-12.7
• US Class Current: 623/25
• CPC Class Codes: A61F 2/72 Bioelectric control, e.g. myoelectric
• CPC Class Codes: A61F 2002/704 computer-controlled, e.g. robotic control

Abstract

A system for controlling the operation of an electrically powered prosthetic appliance which replaces an amputee'"'"'s missing limb. The method employed consists of conducting electromyographic (EMG) signals from the stump of the missing limb and processing the signals so as to reduce them to a near minimum number of linear time series model parameters effective to differentiate between each of the several functions performed by the missing limb from each other function thereof. A filter array carried by the amputee subsequently employs the identified parameters such that each individual filter of the array is tuned according to a set of parameters related to one of the limb functions considered. The myoelectric (EMG or ME) signal as obtained at a surface electrode worn by the amputee is fed to each filter of the above array of filters, and the outputs of these various filters are used for a decision by the system as to which limb function should be actuated. In another realization, identification of partial correlation coefficients of the ME signal is performed. These coefficients are stored in memory per each limb function, and during normal operation the partial correlation parameters for the current ME signal are compared with those in memory to determine which limb function is to be actuated.

Claims

1. In a method for controlling the operation of an electrically powered prosthesis appliance that replaces an amputee'"'"'s missing limb which method includes calibration steps of picking up electromyographic (EMG) signals from the stump of or from at least one muscle close to the point of amputation of the missing limb, and processing the signals thus collected as to reduce same to a near minimum number of linear time-series model parameters effective to differentiate between each of the several functions performed by the missing limb from each other function thereof, and collecting historical data representative of each of the functions of the prosthesis appliance to be controlled for use as background data, the improvement comprising storing the parameters and the background data of each of the functions to be performed by the prosthetic appliance on the person of the amputee, collecting current EMG signals from the stump of the amputee or from at least one muscle close to the point of amputation and passing same simultaneously through several electronic networks having a common input and each having been adjusted by the stored time-series parameters relating to a different limb function as obtained during calibration, and comparing the current EMG signals collected with the outputs of each of said networks, and initiating a response in the prosthetic appliance corresponding to that network whose output over a predetermined time interval is closest to the collected EMG signals over the same time interval.

2. The improved method as set forth in claim 1, wherein the stored background data constitutes processed EMG signals averaged over several runs during which the amputee performs the several functions to be performed by the prosthetic appliance.

3. The improved method as set forth in claim 1, wherein the current EMG signals are repeatedly sampled, processed and compared with the stored historical data during an interval over which said signals remain substantially stationary.

4. The improved method as set forth in claim 1, wherein the stored background data constitutes processed EMG signals averaged over several runs during which the amputee performs the several functions to be performed by the prosthetic appliance, and wherein the historical data thus averaged is stored as sets of parameters (γ

₁₁ . . . γ

_1n);

(γ

₂₁ . . . γ

_2n) . . . (γ

_h1 . . . γ

_hn), where h is the number of the limb functions and n is the number of parameters necessary to discriminate among the several limb functions.

5. The improved method as set forth in claim 4, wherein historical EMG data (y) developed by the amputee and related to time (y_k), where k=1 . . . N, is processed using the stored sets for each limb function in accordance with the following algorithm to yield (y_k,j), where j=1 . . . h

6. The improved method as set forth in claim 5, wherein (_i^min E_i), the minimal E from E₁ to E_n, is compared with stored background data ρ

_i E_i, where ρ

_i is a weight coefficient greater than 1, and wherein if (_i^min E₁) is ≦

ρ

_i E_i the limb function i of the set 1 . . . h is initiated.

7. The improved method as set forth in claim 6, and wherein the limb function is not initiated unless ##EQU9## where A is a predetermined constant and M is the number of y_k² terms.

8. The improved method set forth in claim 7, and wherein if E_i is low due to non-measurement of an EMG signal the prosthesis appliance is placed in a "hold" mode.

9. The improved method of claim 8, and wherein the storing of the background data comprises a first series of calibration runs for deriving and storing the sets of parameters (γ

₁₁ . . . γ

_1n);

(γ

₂₁ . . . γ

_2n) . . . (γ

_h1 . . . γ

_hn) for h different limb functions and a second series of calibration runs for deriving and storing ρ

_i E_i.

10. The improved method of claim 9, and wherein the stored E_i terms are averaged over the second series of calibration runs from the expression ##EQU10## where j=1 . . . h.

11. A method as in claim 1, wherei in the parameters are partial correlation (PC) parameters, and where comparison of current data with stored data relates to measuring distances in some sense between the current set of PC parameters and any set in memory, such that the smallest distance determine the limb function to be actuated.

12. The method as in claim 1, and wherein the steps up to and including storing and data, is performed on a computer in a clinic or laboratory and no on the amputee-borne system.

13. In a system for controlling the operation of an electrically powered prosthetic appliance that replaces an amputee'"'"'s missing limb, which system has been precalibrated by calibration steps of picking up electomyographic (EMG) signals from the stump of or from at least one muscle close to point of amputation of the missing limb, processing the signals thus collected so as to reduce same to a near minimum number of linear time series model parameters effective to differentiate between each of the several functions performed by the missing limb from each other function thereof, and collecting historical data representative of each of the functions of the prosthesis appliance to be controlled for use as background data, the improvement comprising means for storing the parameters and the background data of each of the functions to be performed by the prosthetic appliance on the person of the amputee;

means for collecting current EMG signals from the stump of the amputee or from at least one muscle close to the point of amputation;

a plurality of networks having a common input coupled to receive the collected current EMG signals, each said network having been adjusted by the stored time series parameters relating to respective different limb functions as obtained during calibration;

comparing means for comparing the current EMG signals collected with respective outputs of said networks; and

means responsive to output from said comparing means for initiating a response in the prosthetic appliance corresponding to that network whose output over a predetermined time interval is closest to the collected EMG signals over the same interval.