A machine-learning technique is provided that allows a VR system to automatically learn user gestures based on a ground-truth data set collected from real users. The technique includes two steps. First, ground-truth data is collected by observing multiple users intentionally performing a specified action in a virtual environment. For example, an action to move an object from one place to another is recorded through input from different sensors in the VR system (e.g., position, orientation, controller actuations, or force/acceleration data). Second, machine-learning techniques (e.g., a recurrent neural network, a feedforward neural network, or a Hidden Markov Model) are used to allow the VR system to learn to recognize user gestures intended to represent the actions. The system frees developers from having to custom define each gesture and provides users with accurate responses to natural movements.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.