Gaze Detection Using Facial Symmetry and Pupil Center Shift

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing gaze direction detection technologies require per-subject calibration due to variations in eye shape and pupil contour, making them complex and requiring significant processing.

Innovation Solution

A gaze direction detecting apparatus and method that calculates a reference position for the face and feature positions of the pupils, using a configuration with a reference position calculating section, feature position calculating section, gaze direction feature measure calculating section, and gaze direction calculating section to determine the gaze direction without individual calibration.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If per-subject calibration is performed to account for individual variations in eye shape and pupil contour, then measurement precision of gaze direction is improved, but device complexity and processing requirements increase significantly

Engineering Contradiction:
Improvegaze direction detection accuracyVSAvoidcalibration processing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The invention extracts only the essential geometric relationships needed for gaze detection (pupil center position and line of sight direction) while discarding the need for comprehensive per-subject calibration of eye shape and contour parameters. By focusing on the critical measurement elements rather than all individual variations, the system achieves accurate gaze detection without complex calibration procedures.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The invention creates a universal gaze detection method that works across different subjects without requiring individual calibration. By using standardized geometric calculations based on detected pupil centers and assumed spherical eyeball models, the same detection algorithm can be applied universally to all users, eliminating the need for subject-specific parameter adjustment.

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

2Measurement precision

If comprehensive calibration parameters are collected for each subject, then gaze direction detection accuracy is improved, but processing time and computational load increase

Engineering Contradiction:
Improvegaze direction detection accuracyVSAvoidcalibration processing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The invention extracts only the essential geometric relationships needed for gaze detection (pupil center position and line of sight direction) while discarding the need for comprehensive per-subject calibration of eye shape and contour parameters. By focusing on the critical measurement elements rather than all individual variations, the system achieves accurate gaze detection without complex calibration procedures.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The invention performs preliminary assumptions about eyeball geometry (spherical shape, standard radius) and facial symmetry before actual gaze detection. These pre-established geometric models allow the system to skip time-consuming calibration steps and directly compute gaze direction from detected pupil positions using predetermined calculation methods.

Inventive Principle:
Principle #10Preliminary action

3Measurement precision

If detailed eye contour and pupil boundary analysis is performed, then measurement precision is improved, but adaptability to different individuals decreases due to variability in eye features

Engineering Contradiction:
Improvepupil center detection accuracyVSAvoidapplicability across different subjects
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The invention creates a universal gaze detection method that works across different subjects without requiring individual calibration. By using standardized geometric calculations based on detected pupil centers and assumed spherical eyeball models, the same detection algorithm can be applied universally to all users, eliminating the need for subject-specific parameter adjustment.

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

Solution Approach 2:

The invention changes the detection parameters from detailed eye contour analysis (which varies significantly between individuals) to simplified geometric parameters (pupil center position and line of sight direction) that can be universally measured and calculated. This parameter transformation maintains measurement precision while improving adaptability across different subjects.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentUS8406479B2Visual axis direction detection device and visual line direction detection method
Publication Date: 2013.03.26 PANASONIC INTELLECTUAL PROPERTY CORP OF AMERICA
  • US8406479B2 patent drawing
  • US8406479B2 patent drawing
  • US8406479B2 patent drawing

AI summary

Provided is a visual axis direction detection device capable of obtaining a highly accurate visual axis direction detection result without performing a particular calibration for each of examinees. The device (100) includes: a feature calculation unit (160) which calculates as a reference position, a 3D position of the center of a face of an examinee whose visual axis direction is to be detected, from a 3D position of two face parts positioned symmetrically, and calculates as a feature position, a 3D position of the center of the right and left pupils of the examinee in the horizontal direction; a visual axis direction feature amount calculation unit (172) for calculating a shift amount of the feature position from the reference position in the horizontal direction as a visual axis direction feature amount; and a visual axis vector calculation unit (173) for calculating the visual axis direction of the examinee according to the visual axis direction feature amount.