Eye-Tracker Correction Using Scene Salient Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing eye-tracking technologies face errors due to offsets between the eye camera and the eye's position, which are not effectively corrected by pre-determined markers or three-dimensional information, especially in head-mounted devices.
Innovation Solution
A correction method and device that utilize salient feature information from scene frames captured by a scene camera, analyzed using region of interest or salient object detection methods, to correct eye-tracking operations without pre-determined images or marks, enabling continuous and dynamic correction during use.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If pre-determined markers or three-dimensional information are used for correction, then initial correction can be performed, but the system cannot perform dynamic corrections during wearing and requires removal of the device for re-correction
Solution Approach 1:
The patent transitions from static pre-determined markers to dynamic real-time correction by continuously capturing scene frames and detecting salient objects during device wearing. The correction parameters are dynamically updated based on current scene analysis rather than relying on initial static calibration
Solution Approach 2:
The system performs self-correction by automatically detecting salient objects in the scene and using their known positions to recalibrate the eye-tracking data without requiring external markers or device removal. The device corrects its own tracking errors autonomously during normal operation
2Measurement precision
If pre-determined markers are used for correction, then correction can be performed, but the system requires pre-stored correction images and cannot adapt to different scenes
Solution Approach 1:
The patent extracts correction information directly from the scene content itself by detecting salient objects, rather than relying on pre-stored correction images or external markers. The scene frame contains all necessary information for self-correction
Solution Approach 2:
The salient object detection mechanism serves multiple functions: it identifies objects of interest for the user, provides reference points for correction, and adapts to any scene without requiring pre-stored images. The same detection algorithm works across diverse environments
3Ease of operation
If the eye camera is positioned away from the eye to avoid obstructing the view, then the wearer's view is not blocked, but offset errors are generated in eye-tracking
Solution Approach 1:
The patent introduces salient objects in the scene as intermediary reference points that bridge the gap between the offset eye camera position and the actual eye position. These objects serve as mediators to establish accurate correspondence between captured images and actual gaze directions
Solution Approach 2:
The system continuously captures scene frames, detects salient objects, and uses their known positions to provide feedback for correcting eye-tracking data. This closed-loop feedback mechanism compensates for the offset error caused by the eye camera's distant position
Data Source
AI summary
Correction method and device for an eye-tracker are provided. A non-predetermined scene frame is provided and analyzed to obtain a salient feature information, which is in-turn used to correct an eye-tracking operation. The correction can be done at the initial or during the wearing of an eye-tracker.