Evolving universal gesture sets

摘要

In a gesture-based system, gestures may control aspects of a computing environment or application, where the gestures may be derived from a user's position or movement in a physical space. Gesture recognition data, used to recognize gestures from captured data representative of a user's input gestures, may be evolved based on captured data from a plurality of users. A common set or default set of gesture recognition data may be evolved by selecting a plurality of users for tracking. Captured data of the plurality of users may be processed to identify input gesture data for the plurality of users, and the gesture recognition data may be evolved based on features of the input gesture data that is common to multiple users. The evolved gesture recognition data may be implemented not only for the users tracked, but for users not tracked. An identifier may identify when the evolved gesture recognition data applies and implement the evolved gesture recognition data when the identifier is present.

主权项

1. A method for evolving a gesture set in a gesture-based system, the method comprising:
capturing data representative of a user in a physical space; processing the captured data to identify a first motion or pose for the user, the first motion or pose for the user being processed to determine a first confidence level that the first motion or pose is the same as a computer representation of a motion or pose for a particular gesture that invokes an input command to modify a visual output of a computer, wherein the first motion or pose is processed to generate values for parameters associated with the particular gesture in a gesture filter, said first confidence level indicating a likelihood that the values of the parameters associated with the first motion or pose correspond to the values of the parameters of the particular gesture so as to cause the first motion or pose to be recognized as the particular gesture by the gesture filter to, in turn, invoke the input command to modify the visual output of the computer; and modifying the computer representation of the motion or pose invoking the input command based on the first motion or pose by the user such that subsequently captured data of the user or another user performing the first motion or pose is processed to determine a second confidence level that the first motion or pose is the same as the modified computer representation of the motion or pose for the particular gesture, said second confidence level indicating a likelihood that the values of the parameters associated with the subsequently captured first motion or pose correspond to the values of the parameters of the particular gesture so as to cause the subsequently captured first motion or pose to be recognized as the particular gesture by the gesture filter to, in turn, invoke the input command to modify the visual output of the computer, the second confidence level being greater than the first confidence level.