Motion pattern recognition using wearable motion sensors

1. A computer-implemented method of classifying motion data associated with wearable sensors, comprising:

obtaining, by one or more computing devices, first motion data collected by a first wearable motion sensor during performance of an activity, the first motion data including data associated with one or more first motion primitives comprising classification features identified using machine learning at the first wearable motion sensor;

obtaining, by the one or more computing devices, second motion data collected by a second wearable motion sensor during the performance of the activity, the second motion data including data associated with one or more second motion primitives comprising classification features identified using machine learning at the second wearable motion sensor;

synchronizing, by the one or more computing devices, the first motion data and the second motion data based at least in part on timestamp data associated with the first motion data and the second motion data; and

obtaining, by the one or more computing devices, data associated with a classification of a signature motion pattern associated with the activity, wherein the classification of the signature motion pattern is determined using machine learning at the one or more computing devices and is based at least in part on the one or more first motion primitives and the one or more second motion primitives.

2. The computer-implemented method of claim 1 , further comprising:

providing, at a user interface associated with the one or more computing devices, feedback to a user based on the classification of the signature motion pattern.

3. The computer-implemented method of claim 1 , further comprising:

inputting synchronized motion data into one or more machine-learned models configured to identify signature motion patterns, the synchronized motion data is based on synchronizing the first motion data and the second motion data;

wherein obtaining data associated with the classification of the signature motion pattern comprises obtaining the data associated with the classification from the one or more machine-learned models.

4. The computer-implemented method of claim 3 , wherein:

each of the one or more machine-learned models is associated with a particular activity.

5. The computer-implemented method of claim 3 , wherein:

each of the one or more machine-learned models is associated with a particular type of motion sensor.

6. The computer-implemented method of claim 3 , wherein:

each of the one or more machine-learned models is associated with a particular motion sensor placement.

7. The computer-implemented method of claim 1 , wherein:

the first motion data is collected during performance of a first activity and is associated with a first location on a user body;

the second motion data is collected during performance of the first activity and is associated with a second location on the user body;

the classification of the signature motion pattern is a first classification of a first signature motion pattern; and

the method further comprises:

obtaining third motion data collected by the first wearable motion sensor during performance of a second activity, the third motion data being associated with a third location on the user body and including data associated with one or more third motion primitives; and

obtaining data associated with a second classification of a second signature motion pattern associated with the second activity, the second classification of the second signature motion pattern being determined based at least in part on the one or more third motion primitives.

8. The computer-implemented method of claim 7 , wherein:

obtaining data associated with the first classification of the first signature motion pattern comprises obtaining the data associated with the first classification from a first machine-learned model associated with the first location on the user body; and

obtaining data associated with the second classification of the second signature motion pattern comprises obtaining the data associated with the second classification from a second machine-learned model associated with the second location on the user body.

9. The computer-implemented method of claim 1 , wherein:

the one or more first motion primitives comprise a set of predefined motion parameters.

10. The computer-implemented method of claim 1 , wherein:

the data associated with one or more first motion primitives is identified based at least in part on an analysis of motion patterns representative of signature motions of interest.

11. One or more tangible, non-transitory, computer-readable media that store computer-executable instructions that when executed by one or more processors cause the one or more processors to perform operations, the operations comprising:

obtaining first motion data associated with a first wearable motion sensor and second motion data associated with a second wearable motion sensor, the first motion data including data associated with one or more first motion primitives comprising classification features identified using machine learning at the first wearable motion sensor and corresponding to a first sensor placement relative to a user body, the second motion data including data associated with one or more second motion primitives comprising classification features identified using machine learning at the second wearable motion sensor and corresponding to a second sensor placement relative to the user body;

synchronizing the first motion data and the second motion data based at least in part on one or more times associated with the first motion data and the second motion data;

determining at least one signature motion pattern using machine learning and based at least in part on the first motion data and the second motion data; and

providing at least one output based at least in part on the at least one signature motion pattern.

12. The one or more tangible, non-transitory, computer-readable media of claim 11 , wherein:

determining the at least one signature motion pattern comprises obtaining from a machine-learned model data indicative of a classification of a first signature motion pattern, the classification of the first signature motion pattern determined based at least in part on the one or more first motion primitives and the one or more second motion primitives.

13. The one or more tangible, non-transitory, computer-readable media of claim 12 , wherein:

the machine-learned model is included in a library of machine-learned models; and

each machine-learned model is associated with a particular activity.

14. The one or more tangible, non-transitory, computer-readable media of claim 13 , wherein:

each machine-learned model is associated with at least one of a motion sensor type or a motion sensor placement.

15. The one or more tangible, non-transitory, computer-readable media of claim 14 , wherein the operations further comprise:

training the one or more machine-learned models using the first motion data and the second motion data.

16. The one or more tangible, non-transitory, computer-readable media of claim 11 , wherein:

providing the at least one output comprises providing, at a user interface of one or more computing devices, information associated with the at least one signature motion pattern.

17. The one or more tangible, non-transitory, computer-readable media of claim 16 , wherein:

providing information associated with the at least one signature motion pattern comprises providing data indicative of a comparison of the first motion data and the second motion data relative to one or more reference sets of motion data.

18. The one or more tangible, non-transitory, computer-readable media of claim 16 , wherein providing information associated with the at least one signature motion pattern comprises:

providing, at the user interface, at least one of a visual response, an audible response, or a haptic response.

19. A computing system for classifying motion data associated with wearable sensors, comprising:

one or more processors; and

one or more tangible, non-transitory, computer-readable media that store one or more machine-learned models configured to generate classifications of signature motion patterns based on motion data from a plurality of sensors, the one or more tangible, non-transitory, computer-readable media store instructions that when executed by the one or more processors cause the computing system to perform operations, the operations comprising:

obtaining first motion data collected by a first wearable motion sensor during performance of an activity, the first motion data including data associated with one or more first motion primitives comprising classification features identified using machine learning at the first wearable motion sensor;

obtaining second motion data collected by a second wearable motion sensor during the performance of the activity, the second motion data including data associated with one or more second motion primitives comprising classification features identified using machine learning at the second wearable motion sensor;

synchronizing the first motion data and the second motion data based at least in part on timestamp data associated with the first motion data and the second motion data; and

obtaining from the machine-learned model data associated with a classification of a signature motion pattern associated with the activity, wherein the classification of the signature motion pattern is determined using machine learning at the one or more computing devices and is based at least in part on the one or more first motion primitives and the one or more second motion primitives.