Eye Tracking Calibration Using Scene Camera Object Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current eye tracking calibration methods for head-wearable devices are cumbersome, requiring dedicated environments and trained operators or users, and often result in inaccurate or incomplete data due to user misunderstanding or physical limitations.
Innovation Solution
A method involving a head-wearable device with eye and scene cameras, where a non-translatory movement of an object is displayed, and the user is instructed to mimic this movement, allowing for the generation of datasets that encode gaze-related parameters, enabling accurate calibration without the need for specialized environments or trained users.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If standardized gaze targets are displayed on a screen for calibration, then ground truth gaze positions can be encoded, but the procedure requires a dedicated large display and cannot be performed in an arbitrary environment
Solution Approach 1:
The patent uses the scene camera to capture images of the real-world environment and identifies real-world objects as gaze targets, replacing the need for virtual display targets. This copying of real-world visual information into the calibration process enables environment-independent calibration while maintaining measurement precision.
Solution Approach 2:
The scene camera serves multiple functions: it captures the user's field of view, identifies real-world gaze targets, and provides ground truth position data. This multi-functionality eliminates the need for separate display equipment, enabling calibration in any environment with natural visual stimuli.
2Adaptability or versatility
If a real gaze target is moved by an operator in front of the user, then calibration can be executed everywhere, but it requires a trained operator and a physical gaze target
Solution Approach 1:
The system automatically identifies gaze targets and tracks their positions using the scene camera without requiring operator intervention. The computer vision algorithms autonomously detect objects, calculate their positions, and provide ground truth data, eliminating the need for trained operators and physical targets.
Solution Approach 2:
The patent replaces the mechanical approach of physically moving targets with an optical-digital system. The scene camera captures visual information, and computer vision algorithms process images to track target positions, substituting physical manipulation with automated computational methods.
3Adaptability or versatility
If the user holds a finger and moves the head while gazing at the fingertip, then calibration can be executed everywhere, but it requires a trained user and fingertip detection may not be accurate enough
Solution Approach 1:
Instead of using the user's finger as a physical target, the system uses the scene camera to capture and identify stable real-world objects in the environment. These identified objects serve as virtual gaze targets, providing consistent and accurately detectable features for calibration.
Solution Approach 2:
The patent changes the target detection parameter from tracking human body parts (fingertip) to tracking environmental objects with distinct visual features. This parameter change improves detection accuracy by utilizing objects with more stable and distinguishable characteristics in the scene camera images.
4Measurement precision
If the user moves the head with respect to a device held by the user while keeping the gaze on a static marker, then detection accuracy improves, but the user still has no guidance about what to do during the procedure
Solution Approach 1:
The system provides real-time feedback by displaying the identified gaze target and its position to the user during calibration. This visual feedback guides the user on where to look and how to move their head, making the procedure intuitive while maintaining accurate target detection.
Solution Approach 2:
The system acts as an intermediary by automatically identifying and presenting gaze targets to the user. This intermediary function bridges the gap between the user's natural viewing behavior and the calibration requirements, providing clear guidance without requiring detailed instructions.
Data Source
AI summary
A system and method for generating data suitable for calibrating a head-wearable device is disclosed. In one example the device includes a first eye camera and a scene camera. The first eye camera is used to generate first images of at least a portion of a first eye of the user while the user is expected to look at the object and mimic the not translatory movement. Respective positions of the object in the field images are determined. The determined positions of the object are used to determine the first images respective ground truth values of at least one gaze-direction related parameter of the user.