Multi-User Gaze Tracking for Accurate 3D Display Rendering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D display technologies struggle to provide accurate and efficient image rendering for multiple viewers, often projecting distorted or broken images due to inadequate viewer positioning and gaze tracking, leading to increased latency and poor user experience.
Innovation Solution
Implementing deep learning systems for face detection, gaze tracking, and camera image evaluation to determine viewer-specific 3D projections by analyzing face and eye landmarks, head pose, and gaze direction, ensuring appropriate image rendering for each viewer.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If gaze tracking is implemented for multiple viewers, then viewer-specific 3D image rendering accuracy is improved, but system complexity and processing time increase
Solution Approach 1:
The system segments the gaze tracking process into independent modules: face detection, eye landmark detection, head pose estimation, and gaze direction calculation. Each module processes specific data independently, allowing parallel computation and reducing overall system complexity while maintaining high accuracy for multiple viewers
Solution Approach 2:
The patent introduces an intermediary processing layer that uses pre-computed face landmarks and head pose information as intermediaries to calculate gaze direction. This intermediary approach simplifies the complex direct gaze calculation by breaking it down into manageable steps using intermediate geometric relationships
2Speed
If real-time gaze tracking is performed for multiple viewers, then rendering responsiveness is improved, but processing latency increases
Solution Approach 1:
The system performs preliminary actions by pre-detecting face landmarks and pre-estimating head pose in earlier processing stages. These pre-computed values are then used in subsequent gaze direction calculations, reducing the computational burden during real-time rendering and minimizing latency
Solution Approach 2:
The patent implements optimization that skips unnecessary computational steps by using geometric constraints and pre-computed data. When certain conditions are met (e.g., stable head pose), the system can directly calculate gaze direction without re-processing all previous steps, thereby reducing processing time while maintaining accuracy
3Manufacturing precision
If comprehensive eye tracking parameters are collected, then 3D projection accuracy is improved, but data processing requirements increase
Solution Approach 1:
The system applies local quality by collecting and processing only the specific eye tracking parameters that are locally relevant to each viewer's position and gaze direction. Not all possible eye parameters are processed for every viewer; instead, the system selectively processes parameters based on the specific rendering requirements of each viewer, improving processing efficiency
Solution Approach 2:
The patent dynamically changes the set of processed parameters based on viewing conditions. When a viewer is clearly positioned and their gaze is stable, the system uses fewer parameters. When positioning is uncertain or gaze is dynamic, the system activates additional parameters for more accurate tracking, thereby optimizing processing efficiency for each specific scenario
Data Source
AI summary
Methods, systems, and storage media for projecting multi-viewer-specific 3D object perspectives from a single 3D display are disclosed. Implementations may: acquire face and eye region image data of a plurality of viewers within a field of view of at least one camera associated with a 3D-enabled digital display; analyze the eye region image data to determine at least one 3D eye position, at least one eye state, at least one gaze angle, and at least one point-of-regard for at least one viewer relative to at least one camera associated with the 3D-enabled digital display; and calculate a plurality of image projections for display by the single 3D display.