A Machine Learning Approach to Expanding the Degrees of Freedom on Phone-Based Head Mounted Displays

Phone-based head mounted displays (HMDs) make participation in virtual reality (VR) easily accessible. However, phone-based HMDs can only track the user’s head orientation, excluding head position. This not only stifles the user’s immersion in the virtual environment, but also risks simulator sickness. To mitigate this limitation, we propose a neural network that predicts the user’s head position given the head orientation history and other information about the user’s movement that can be tracked from phone sensors. To analyze performance, the model iterates over the test dataset, predicting the entirety of the dataset’s positional displacement history.