Head Pose Prediction Using High-Order Motion Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing terminal devices inaccurately predict user head pose when movement is vigorous due to reliance on constant velocity assumptions, leading to poor accuracy in pose prediction.
Innovation Solution
Acquire first motion information during a historical period, determine high-order information, and use it to predict second motion information for future periods, incorporating current pose to accurately forecast target poses using high-order coefficient matrices and inverse matrices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If constant velocity assumption is used for pose prediction, then calculation complexity is reduced, but prediction accuracy deteriorates when user moves vigorously
Solution Approach 1:
The patent transitions from a static constant velocity model to a dynamic high-order motion model. By introducing high-order information (acceleration and higher derivatives) that changes over time, the system can adapt to varying motion patterns of the user's head, enabling accurate prediction during vigorous movement while maintaining computational efficiency through structured mathematical approaches.
Solution Approach 2:
The patent changes the fundamental parameters of motion modeling from first-order (constant velocity) to high-order (variable acceleration and beyond). This parameter transformation allows the system to capture complex motion dynamics by incorporating time-varying acceleration, velocity, and position relationships, thereby improving prediction accuracy without proportionally increasing complexity.
2Measurement precision
If high-order motion information is incorporated, then pose prediction accuracy is improved, but calculation complexity increases
Solution Approach 1:
The patent performs preliminary computation by pre-calculating and storing motion information at multiple time steps before the prediction moment. By preparing acceleration, velocity, and position data in advance during the historical period, the system reduces the computational burden during the actual prediction phase, making the complex high-order modeling manageable while maintaining accuracy.
Solution Approach 2:
The patent maintains continuous computation of motion parameters throughout the historical period, continuously updating acceleration, velocity, and position information. This continuous action ensures that the most recent and relevant motion data is always available for prediction, improving accuracy by incorporating up-to-date motion patterns without requiring complex batch processing.
Data Source
AI summary
The present disclosure provides a pose prediction method and apparatus, a terminal device, and a storage medium, and the method includes: acquiring first motion information of a user during a first historical period and a current pose of the user; determining high-order information associated with motion of the user during the first historical period according to the first motion information; determining second motion information of the user during a future period according to the high-order information and the first motion information; and predicting a target pose of the user at a target moment according to the second motion information of the user during the future period and the current pose.