Lateral Gaze Angle Estimation via Eye Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting direct gaze, such as in self-service kiosks and driver alertness monitoring, are inadequate as they either fail to ensure the user is looking at the device or rely on computationally complex systems that are not practical for real-time applications, especially when head rotation and gaze direction are involved.
Innovation Solution
The method estimates lateral gaze angle by measuring the relative eye separation using video cameras, determining the distance between the pupils, and applying a simple computational formula to infer the gaze angle, which can be used to infer the user's interest in their environment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional gaze detection methods (neural net classifier, dense stereo reconstruction) are used, then measurement precision is improved, but device complexity and computational cost increase significantly
Solution Approach 1:
The patent extracts only the essential feature needed for gaze detection - the distance between the two eyes - from the complex facial structure. By focusing solely on inter-ocular distance rather than analyzing entire facial images or performing dense stereo reconstruction, the system achieves adequate measurement precision with dramatically reduced computational complexity
Solution Approach 2:
Instead of using complex algorithms to directly compute gaze angle from image data, the patent inverts the approach by using a simple geometric relationship: the inter-ocular distance in the image plane can be used to infer the rotation angle of the head. This inversion transforms a complex computer vision problem into a simple geometric calculation
2Device complexity
If simple proximity sensors or background subtraction are used, then device complexity is reduced, but measurement precision deteriorates as they cannot distinguish direct gaze from mere proximity
Solution Approach 1:
The patent moves from two-dimensional image plane analysis to three-dimensional spatial reasoning by utilizing the known fixed distance between the user's eyes. By incorporating this depth dimension (the inter-ocular distance in 3D space), the system can accurately determine gaze direction using simple 2D image measurements combined with this third-dimensional constraint
Data Source
AI summary
This invention provides for the automatic determination of a human's gaze angle. The gaze angle is assumed to be the same as the head angle (i.e., eyes looking straight ahead). It is determined by comparing the observed eye separation relative to a maximal eye separation. The gaze angle is then determined to be the inverse cosine, or arc cosine, of the ratio of these two measurements.