Deep predictor recurrent neural network for head posture prediction

摘要

A system and method for predicting head posture for an extended or virtual reality device rendering engine can include a recurrent neural network (RNN) that receives time series head posture data and outputs a predicted head pose.The recurrent neural network may include one or more long - term memory (LSTM) units or gated recursive units (Gru).The full coupling (FC) layer receives an input from the RNN and can output a 3 DOF (DOF) head posture (e.g., angular orientation or spatial position) or 6 DOF head posture (e.g., both angular orientation and spatial position).The rendering engine can generate virtual content and display it to the user using the predicted head posture when the user is looking toward the location of the virtual content, which reduces system latency and improves user experience.