Determining intended user movement to control an external device

1. A method, comprising:

receiving, by a system comprising a processor, user-specific EMG signals recorded by electrodes located on a limb of a user, wherein the user-specific EMG signals comprise one or more EMG patterns that indicate a single movement;

decomposing, by the system, the user-specific EMG signals into the one or more EMG patterns in EMG feature space that indicate the single movement;

updating, by the system, a user-specific model of muscular activity to include the single movement and corresponding one or more primary muscle patterns based on the one or more EMG patterns in EMG feature space; and

controlling, by the system, the external device based on a predicted intent of the user to move the external device using the user-specific model of muscular activity by:

receiving an unknown EMG signal from the user,

identifying a partition of the EMG feature space where the unknown EMG signal is located,

bounding the partition of the EMG feature space where the unknown EMG signal is located by known single movements with a corresponding one or more primary muscle patterns, and

determining a ratio of movement vectors that should be combined to determine the direction and the magnitude of the intent of the user to move the external device, wherein the ratio of movement vectors corresponds to the unknown EMG signal.

2. The method of claim 1 , wherein the receiving the user-generated EMG signals further comprises:

displaying a visualization of a number of combinations of movements; and

prompting the user to move the limb of the user to match the visualization.

3. The method of claim 2 , wherein the movements comprise wrist flexion/extension, wrist pronation/supination, D2 flexion/extension, thumb palmer movement, thumb lateral movement, wrist radial/ulnar deviation, and D3-D5 flexion/extension.

4. The method of claim 1 , wherein the updating the user-specific model further comprises:

extracting a feature from each the one or more EMG patterns in EMG feature space that indicate the single movement, wherein each of the user specific EMG signals is recorded by a different electrode;

determining, by a fitting method, the one or more primary muscle patterns for the feature;

generating the EMG feature space based on the feature extracted for the single movement;

partitioning the EMG feature space into a tetrahedra of N+1 vertices; and

generating a Triangulated Irregular Network based on the tetrahedra of N+1 vertices, wherein the Triangulated Irregular Network corresponds to the user-specific model of muscular activity.

5. The method of claim 4 , wherein the EMG feature space is partitioned with Delaunay Triangulation, wherein the EMG feature space is divided into the tetrahedra with a maximized minimum internal angle.

6. The method of claim 4 , wherein the feature is a mean absolute value.

7. The method of claim 4 , wherein the primary muscle patterns for the feature are determined by Principal Component Analysis.

8. The method of claim 4 , wherein the one or more primary muscle patterns are fixed ratios of muscle synergies.

9. The method of claim 1 , wherein the electrodes are implanted in at least two muscles of the limb of the user.

10. The method of claim 9 , wherein the at least two muscles comprise at least two of a Pronator, a Flexor Carpi Radialis (FCR), a Flexor Digitorum Superficialis (FDS), a Flexor Carpi Ulnaris (FCU), a Supinator, an Extensor Carpi Radialis Longus (ECRL), an Extensor Digitorum (ED), or an Extensor Carpi Ulnaris (ECU).

11. The method of claim 1 , wherein the external device is one of a prosthetic limb, a robotic system, or a virtual system.

12. A system, comprising:

at least one electrode configured to be located on a limb of a user; and

a controller, coupled to the at least one electrode, comprising:

a non-transitory memory storing instructions; and

a processor to execute the instruction stored in the memory to at least:

receive user-specific EMG signals recorded by electrodes located on a limb of a user, wherein the user-specific EMG signals comprise one or more EMG patterns that indicate a single movement;

decompose the user-specific EMG signals into the one or more EMG patterns in EMG feature space that indicate the single movement;

update a user-specific model of muscular activity to include the single movement and corresponding one or more primary muscle patterns based on the one or more EMG patterns in EMG feature space; and

predict an intent of the user to move the external device based on the user-specific model of muscular activity by:

identifying a partition of the EMG feature space where the unknown EMG signal is located;

bounding the partition of the EMG feature space where the unknown EMG signal is located by known single movements with a corresponding one or more primary muscle patterns; and

determining a ratio of movement vectors that should be combined to determine the direction and the magnitude of the intent of the user to move the external device, wherein the ratio of movement vectors corresponds to the unknown EMG signal,

wherein the user-specific model of muscular activity is used to control an external device based on muscular activity in the limb of the user.

13. The system of claim 12 , wherein the system further comprises:

a display device coupled to the controller and configured to:

display a visualization of a number of combinations of movements; and

prompt the user to move the limb of the user to match the visualization.

14. The system of claim 12 , wherein the user-specific model is updated by:

extracting a feature from each the one or more EMG patterns in EMG feature space that indicate the single movement, wherein each of the user specific EMG signals is recorded by a different electrode;

determining, by a fitting method, the one or more primary muscle patterns for the feature;

generating the EMG feature space based on the feature extracted for the single movement;

partitioning the EMG feature space into a tetrahedra of N+1 vertices; and

generating a Triangulated Irregular Network based on the tetrahedra of N+1 vertices, wherein the Triangulated Irregular Network corresponds to the user-specific model of muscular activity.

15. The system of claim 14 , wherein the EMG feature space is partitioned with Delaunay Triangulation, wherein the EMG feature space is divided into the tetrahedra with a maximized minimum internal angle.

16. The system of claim 14 , wherein the feature is a mean absolute value.

17. The system of claim 14 , wherein the primary muscle patterns for the feature are determined by Principal Component Analysis.

18. The system of claim 12 , wherein the electrodes are implanted in at least two muscles of the limb of the user.

19. The system of claim 18 , wherein the at least two muscles comprise at least two of a Pronator, a Flexor Carpi Radialis (FCR), a Flexor Digitorum Superficialis (FDS), a Flexor Carpi Ulnaris (FCU), a Supinator, an Extensor Carpi Radialis Longus (ECRL), an Extensor Digitorum (ED) or an Extensor Carpi Ulnaris (ECU).

20. The system of claim 12 , wherein the external device is one of a prosthetic limb, a robotic system, or a virtual system.

21. A method, comprising:

receiving, by a system comprising a processor, user-specific EMG signals recorded by electrodes located on a limb of a user, wherein the user-specific EMG signals comprise one or more EMG patterns that indicate a single movement;

decomposing, by the system, the user-specific EMG signals into the one or more EMG patterns in EMG feature space that indicate the single movement; and

updating, by the system, a user-specific model of muscular activity to include the single movement and corresponding one or more primary muscle patterns by:

extracting a feature from each the one or more EMG patterns in EMG feature space that indicate the single movement, wherein each of the user specific EMG signals is recorded by a different electrode;

determining, by a fitting method, the one or more primary muscle patterns for the feature;

generating the EMG feature space based on the feature extracted for the single movement;

partitioning the EMG feature space into a tetrahedra of N+1 vertices; and

generating a Triangulated Irregular Network based on the tetrahedra of N+1 vertices, wherein the Triangulated Irregular Network corresponds to the user-specific model of muscular activity,

wherein the user-specific model of muscular activity is used to control an external device based on muscular activity in the limb of the user.