Gaze Estimation via Common Coordinate System Calibration

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing gaze estimation methods require cumbersome calibration procedures, leading to relatively inaccurate estimations, especially when passively tracking user attention in scenes like shopping windows, as they rely on individual-specific or non-scene-specific statistical data.

Innovation Solution

A system and method that calibrate un-calibrated eye measurement points from different users viewing the same scene by mapping them to a common coordinate system, utilizing the insight that gaze patterns between users are similar for the same scene, thereby establishing a more accurate scene transformation for estimating gaze points without individual calibration.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If calibration procedures are performed to improve gaze estimation accuracy, then measurement precision improves, but device complexity and time consumption increase

Engineering Contradiction:
Improvegaze estimation accuracyVSAvoidcalibration procedure complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system performs automatic self-calibration by capturing natural eye movement data during normal viewing and using statistical analysis to determine calibration parameters without requiring active user participation or manual adjustment procedures

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system changes the approach from requiring users to fixate on specific calibration points to using statistical distribution of natural eye movement measurements to derive calibration parameters, fundamentally altering the calibration process parameters

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If individual-specific calibration data is used to improve gaze estimation accuracy, then measurement precision improves, but loss of time increases due to per-user calibration requirements

Engineering Contradiction:
Improvegaze estimation accuracyVSAvoidcalibration time per user
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system collects and analyzes eye movement data from multiple users in advance to establish statistical models and calibration parameters that can be applied to new users without requiring them to undergo time-consuming calibration procedures

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system creates universal calibration models based on statistical analysis of eye movement patterns across multiple users, making the calibration data applicable to multiple users rather than being strictly individual-specific

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Ease of operation

If passive gaze tracking is implemented to reduce user participation, then ease of operation improves, but measurement precision deteriorates

Engineering Contradiction:
Improvepassive tracking capabilityVSAvoidgaze estimation accuracy
Core Design Contradiction:
Ease of operationVSMeasurement precision

Solution Approach 1:

The system uses statistical feedback from aggregated eye movement data of multiple users to continuously refine and improve the accuracy of passive gaze estimation, using the collected data to adjust and optimize calibration parameters

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP3074844B1Estimating gaze from un-calibrated eye measurement points
Publication Date: 2018.03.28 SIGHTCORP
  • EP3074844B1 patent drawingFigure 1~3
  • EP3074844B1 patent drawingFigure 4~5
  • EP3074844B1 patent drawingFigure 6

AI summary

The invention enables estimating gaze from a set of eye measurement points which are indicative of a gaze pattern of a user viewing a scene. Herein, use is made of the insight that gaze patterns of different users are similar for a same scene (010). The invention involves obtaining different sets of eye measurement points from different users viewing the same scene. The different sets of eye measurement points may be mapped to a common coordinate system, and there by mutually calibrated, based on the assumption that geometrically similar topologies between the different sets of eye measurement points are to be mapped to similar coordinates in the common coordinate system as they represent a similar or same gaze pattern.A scene transformation for mapping the common coordinate system to a coordinate system associated with the scene can be calculated by matching eye measurement points from the common coordinate system to interest points(012) of the scene. The scene transformation is thereby calculated more accurately than individually calculated scene transformations, thereby providing a more accurate estimate of the gaze points.