SYSTEM AND METHOD FOR GENERATING A VIRTUAL AVATAR

1 . A non-transitory computer-readable medium storing computer-executable instructions, which, when executed by a computer processor, cause the computer processor to carry out a method of generating a virtual avatar representing a user using sensor data from a plurality of sensors mounted to the user wherein the method comprises:

obtaining sensor data from the plurality of sensors, wherein the sensor data includes:

force sensor data from a plurality of force sensors positioned underfoot of the user; and

IMU data from a plurality of inertial measurement units (IMUs) mounted at respective IMU locations on the user, wherein the IMU data includes a plurality of IMU datasets, and each IMU dataset corresponds to one of the IMU locations;

determining an activity type from the sensor data;

determining an estimated full-body posture of the user by inputting the IMU data, force sensor data, and activity type to a machine learning model trained to estimate full-body posture, wherein the estimated full-body posture specifies a limb position and a limb orientation for each limb of the user;

generating the virtual avatar based on the estimated full-body posture; and

displaying the virtual avatar.

2 . The non-transitory computer-readable medium of claim 1 , wherein generating the virtual avatar comprises:

determining a plurality of skeletal lines representing the limbs of the user based on the estimated full-body posture; and

generating the virtual avatar from the plurality of skeletal lines.

3 . The non-transitory computer-readable medium of claim 2 , wherein generating the virtual avatar from the plurality of skeletal lines comprises:

generating an avatar base template from the plurality of skeletal lines; and

scaling the avatar base template using anthropometric measurement data for the user to generate the virtual avatar.

4 . The non-transitory computer-readable medium of claim 1 , further comprising determining motion of the limbs of the user based on limb velocity data and acceleration data from the IMU data.

5 . The non-transitory computer-readable medium of claim 4 , further comprising moving the virtual avatar within a virtual environment in response to the motion of the limbs.

6 . The non-transitory computer-readable medium of claim 1 , wherein the plurality of IMUs includes a pair of foot-mounted IMUs, and the activity type is determined based on the IMU datasets from the pair of foot-mounted IMUs.

7 . The non-transitory computer-readable medium of claim 1 , wherein the plurality of IMUs comprises at most 10 IMUs.

8 . The non-transitory computer-readable medium of claim 1 , wherein the plurality of IMUs comprises a pair of underfoot IMUs, a pair of wrist-mounted IMUs, a sacrum-mounted IMU, a trunk-mounted IMU, a pair of thigh-mounted IMUs, and a pair of shank-mounted IMUs.

9 . The non-transitory computer-readable medium of claim 1 , wherein the virtual avatar is generated in the absence of any visual motion capture data.

10 . The non-transitory computer-readable medium of claim 1 , wherein the machine learning model is trained using training data generated by collecting training sensor data and training motion capture data and calculating a training full-body posture using the training motion capture data.

11 . The non-transitory computer-readable medium of claim 10 , wherein the training sensor data and training motion capture data is collected from one or more users while performing a plurality of specified activity types.

12 . The non-transitory computer-readable medium of claim 10 , wherein:

the training motion capture data from a given training user is used to generate an initial training virtual avatar; and

the initial training virtual avatar is scaled using anthropometric measurement data for the given training user to generate a training virtual avatar.

13 . The non-transitory computer-readable medium of claim 10 , wherein the machine learning model is trained by:

calculating a training estimated full-body posture by inputting the training sensor data to the machine learning model; and

optimizing the machine learning model based on a correlation between the training estimated full-body posture and the training full-body posture.

14 . The non-transitory computer-readable medium of claim 3 , wherein a first scaling factor applied to a motion of the virtual avatar and a second scaling factor applied to a size of the virtual avatar are different.

15 . The non-transitory computer-readable medium of claim 1 , wherein generating the virtual avatar and displaying the virtual avatar are performed in real time, substantially coincident with obtaining the sensor data.

16 . The non-transitory computer-readable medium of claim 1 , wherein the virtual avatar is displayed on a VR headset or other display.

17 . The non-transitory computer-readable medium of claim 1 , wherein the virtual avatar is used to visualize differences in performance or to be used as a tool for reducing a risk of injury.

18 . The non-transitory computer-readable medium of claim 1 , wherein the activity type includes walking, running, cycling, jumping, dancing, golfing, snowboarding, or other athletic activity.

19 . The non-transitory computer-readable medium of claim 1 , further comprising comparing virtual avatars from different activity sessions to visualize differences in performance.

20 . The non-transitory computer-readable medium of claim 4 , wherein virtual avatar motion is proportional to the motion of the limbs of the user.