Avatar Eye Gaze Encoding for 3D Look-At Animation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current scene description formats lack standardized methods for encoding and transmitting 'look-at' gaze information for avatars, limiting the ability of rendering applications to seamlessly animate eyes or comparable objects in virtual 3D spaces, as they often require prior knowledge and do not allow independent rotation of eyes to focus on specific locations.

Innovation Solution

A method and device for generating and decoding 3D scenes with avatars that include gaze information, using a scene graph to encode angular and look-at gaze types with parameters for horizontal, vertical, and roll rotations, along with gaze style and speed, enabling independent eye rotation and dynamic gaze changes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If standardized gaze information is encoded in scene descriptions, then rendering applications can seamlessly animate eyes to focus on specific points in 3D space, but the complexity of the scene description format increases

Engineering Contradiction:
Improveease of animating avatar gazeVSAvoidcomplexity of scene description format
Core Design Contradiction:
Ease of operationVSDevice complexity

Solution Approach 1:

The gaze information is segmented into distinct components: horizontal rotation, vertical rotation, and roll rotation. Each component is encoded separately in the scene description, allowing the rendering application to process and apply them independently to achieve realistic eye movement without overwhelming complexity

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The gaze information including rotation parameters and animation metadata is prepared and encoded in advance within the scene description file. This preliminary encoding allows the rendering application to directly use the pre-processed data without requiring complex real-time calculations, simplifying the runtime operation

Inventive Principle:
Principle #10Preliminary action

2Manufacturing precision

If independent eye rotation is enabled for each eye, then avatars can focus on specific close points in 3D space, but the number of parameters and data required increases

Engineering Contradiction:
Improveprecision of gaze directionVSAvoidquantity of gaze parameters
Core Design Contradiction:
Manufacturing precisionVSQuantity of substance

Solution Approach 1:

Each eye is assigned its own independent rotation parameters (horizontal, vertical, roll) tailored to its specific position and orientation on the avatar's head. This local quality approach allows precise control of each eye's movement while the parameters are efficiently encoded only where needed, avoiding unnecessary data duplication

Inventive Principle:
Principle #3Local quality

3Adaptability or versatility

If angular gaze encoding is used, then avatar rotation can be controlled along three axes, but both eyes cannot point to specific close locations independently

Engineering Contradiction:
Improveversatility of gaze controlVSAvoidprecision of eye positioning
Core Design Contradiction:
Adaptability or versatilityVSManufacturing precision

Solution Approach 1:

The system transitions from static angular encoding to dynamic independent control of each eye. Each eye can now rotate independently around its local axes, allowing them to converge on close points or diverge as needed. This dynamic approach maintains the versatility of three-axis rotation while adding the precision needed for independent eye positioning

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP4676070A1Avatar "look at" gaze in scene descriptions
Publication Date: 2026.01.07 INTERDIGITAL CE PATENT HOLDINGS SAS
  • EP4676070A1 patent drawingFigure 1~2
  • EP4676070A1 patent drawingFigure 3~4
  • EP4676070A1 patent drawingFigure 5

AI summary

Methods, apparatus and data stream are provided encoding, transmitting and decoding of the "look at" gaze of an avatar. The generated scene description comprises information for getting the avatar looking at a given point in the virtual three-dimensional (3D) space. Different animations and related parameters are described in a generic format for each eye (or equivalent object's subcomponent) that allows to finely animate the eye(s) of an avatar, in particular to permit a "look at a point" animation type.