Multi-Camera Eye Gaze Tracking with Confidence Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing eye gaze tracking systems struggle to accurately track eye movement when a user's head moves, especially if their eyes are partially obscured or in changing light conditions, and they often fail to function effectively for users wearing glasses or under sudden changes in lighting.
Innovation Solution
A system utilizing multiple light sources and image sensors, with a gaze tracking module that includes ocular feature extraction, point of regard (POR) calculation, and averaging modules, which processes gaze data to determine a confidence value for each image sensor, allowing for robust tracking by weighting accurate data and compensating for head pose, occlusion, and varying light conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single camera is used for eye gaze tracking, then the system is simple and inexpensive, but the system cannot accurately track gaze when the user moves their head or when eyes are partially obscured
Solution Approach 1:
The system divides the gaze tracking task across multiple cameras positioned at different locations. Each camera captures eye images from its own perspective, and the system segments the processing by having each camera independently calculate POR with associated confidence values, which are then combined to produce a robust final gaze position.
Solution Approach 2:
The system merges data from multiple cameras by combining their individual POR calculations and confidence values. This integration allows the system to maintain accurate gaze tracking even when one camera's view is blocked or when the user moves their head, as other cameras can compensate.
2Reliability
If multiple image sensors are used to improve tracking robustness, then the system can handle head movements and occlusions, but the computational overhead increases
Solution Approach 1:
The system applies local quality assessment by calculating confidence values for each camera's POR measurement based on local conditions such as head pose angle, feature detection reliability, and whether the eye is obscured. This allows the system to weight each camera's contribution appropriately without uniformly processing all data at full complexity.
Solution Approach 2:
The system changes parameters dynamically by adjusting confidence values based on real-time conditions. When a camera detects poor conditions (e.g., large head pose angle, obscured eye, or low feature detection reliability), it reduces the confidence value for that camera's POR measurement, effectively down-weighting its contribution to the final result without requiring complex reprocessing.
3Ease of operation
If traditional optical tracking systems are used, then the system is non-invasive and inexpensive, but the system fails when eyes are close to or outside the camera's visual field
Solution Approach 1:
The system transitions from a single 2D camera view to a multi-dimensional spatial arrangement with cameras positioned at different locations around the display. This dimensional expansion creates multiple visual fields that collectively cover a larger area, allowing eyes to remain within at least one camera's field of view even during head movements.
Solution Approach 2:
The system achieves universality by designing a multi-camera configuration that handles multiple scenarios simultaneously - normal viewing, head movements, partial occlusions, and varying light conditions. Each camera serves multiple functions by contributing to gaze tracking under different conditions, making the system universally robust across diverse usage situations.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system can accurately and robustly track eye gaze even when the eyes are close to or outside the visual field, partially obscured, or in changing light conditions, providing efficient and economical operation with reduced computational overhead.
Implementation Method 1
a gaze data acquisition system comprising a plurality of light sources and a plurality of image sensors, the plurality of light sources arranged to emit light to a head of the user, and the plurality of image sensors configured to receive light from the head of the user to capture gaze data thereof
Data Source
AI summary
An eye gaze tracking system is disclosed. The system includes a gaze data acquisition system including a plurality of light sources and a plurality of image sensors. The plurality of light sources are arranged to emit light to a head of the user, and the plurality of image sensors are configured to receive the light. In an embodiment, the system further includes a gaze tracking module including an ocular feature extraction module, a point of regard (POR) calculation module and a POR averaging module. The ocular feature extraction module is configured to process the gaze data and to extract ocular features, and is configured to determine a confidence value associated with an accuracy of the parameters. The POR calculation module is configured to determine a POR from the ocular features. The POR averaging module is configured to determine an average POR.