Virtual reality presentation of body postures of avatars

1. A method implemented in a three-dimensional virtual reality world, the method comprising:

receiving input data tracking a position, orientation, and motions of a head of a user of the virtual reality world, wherein an avatar represents the user in the virtual reality world, wherein the input data is generated by a virtual reality (VR) headset;

predicting, from the input data, a posture of the avatar according to a posture model,

wherein the posture is predicted based at least in part on a context of the avatar in the virtual reality world;

identifying a pattern of head motion of the user from the input data generated by the VR headset, wherein the prediction is based on a detection of the pattern in the input data, and wherein the pattern represents a head gesture of the user;

determining personalization parameters of the posture model based on the head gesture and using a machine learning technique, wherein the posture is predicted based on personalization parameters of the avatar; and

computing an animation of the avatar showing the posture of the avatar in the virtual reality world.

2. The method of claim 1 , wherein the posture identifies a position change or movement of a body part of the avatar relative to a head of the avatar.

3. The method of claim 2 , wherein the prediction is made without an input device sensing a position or movement of a portion of the user corresponding to the body part of the avatar.

4. The method of claim 3 , wherein the body part includes a limb or torso of the avatar.

5. The method of claim 4 , wherein the avatar is presented in the virtual reality world to second avatars in the virtual reality world.

6. The method of claim 5 , wherein the posture of the avatar is not presented to the user in the virtual reality world.

7. The method of claim 1 , further comprising:

controlling an orientation and position of the avatar based on the input data.

8. The method of claim 1 , wherein the posture model is trained via machine learning using postures captured in video images of people engaging in social activities.

9. The method of claim 1 , wherein the posture model is trained via machine learning using body movements tracked using position and motion tracking devices attached to a portion of users of the virtual reality world.

10. The method of claim 1 , wherein the context comprises a context selected from the group consisting of listening to a speech, engaging in a conversation, giving a speech, and hearing a greeting or alert.

11. A non-transitory computer storage medium storing instructions configured to instruct a computer device to perform a method implemented in a three-dimensional virtual reality world, the method comprising:

receiving input data tracking a position, orientation, and motions of a head of a user of the virtual reality world, wherein an avatar represents the user in the virtual reality world, wherein the input data is generated by a virtual reality (VR) headset;

predicting, from the input data, a posture of the avatar according to a posture model,

wherein the posture is predicted based at least in part on a context of the avatar in the virtual reality world;

identifying a pattern of head motion of the user from the input data generated by the VR headset, wherein the prediction is based on a detection of the pattern in the input data, and wherein the pattern represents a head gesture of the user;

determining personalization parameters of the posture model based on the head gesture and using a machine learning technique, wherein the posture is predicted based on personalization parameters of the avatar; and

computing an animation of the avatar showing the posture of the avatar in the virtual reality world.

12. A computing system to implement a three-dimensional virtual reality world, the system comprising:

a server system; and

a data storage device storing:

a three-dimensional model of the virtual reality world; and

avatar models representing residences of the virtual reality world;

wherein the server system generates, from the three-dimensional model of the virtual reality world and the avatar models, data stream to provide views of the virtual reality world to client devices that are connected to the server system via a computer network;

wherein the computing system tracks a position and orientation of a respective avatar in the virtual reality world;

wherein in response to input data, generated by a virtual reality (VR) headset, tracking a position, orientation, and motions of a head of a user of the virtual reality world, the server system uses a posture model to:

predict, from the input data, a posture of an avatar of the user in the virtual reality world;

identify a pattern of head motion of the user from the input data generated by the VR headset, wherein the prediction is based on a detection of the pattern in the input data, and wherein the pattern represents a head gesture of the user;

determine personalization parameters of the posture model based on the head gesture and using a machine learning technique, wherein the posture is predicted based on personalization parameters of the avatar; and

compute an animation of the avatar showing the posture of the avatar in the virtual reality world; and

wherein the posture is predicted based at least in part on a context of the avatar in the virtual reality world.

13. The computing system of claim 12 , wherein the animation is computed without input from a client device tracking relative positions or movements of portions of the user that corresponding to the relative positions or movements of corresponding portions of the avatar having the posture.