Generative Neural Network Anonymizing Video Identities

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing technologies face challenges in anonymizing identities in videos or images while preserving behavior information, making it difficult to automate the process of predicting emotional or internal cognitive states without revealing personal data.

Innovation Solution

A computing system utilizing a generative neural network is configured to receive input videos with distinct identities and similar behaviors, synthesizing output videos with a synthetic face and identity that differs from the input videos, while preserving the behavior information using a loss function that balances similarity, behavior preservation, and face consistency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If statistical machine learning methods are used to predict emotional or internal cognitive states, then prediction accuracy is improved, but large amounts of personal data must be stored

Engineering Contradiction:

Improveprediction accuracyVSAvoidamount of personal data

Core Design Contradiction:

Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent extracts and separates identity information from behavior information through the encoder network. The encoder extracts behavior-related features while deliberately excluding identity-specific features, allowing behavior analysis without storing or processing personal identity data. This extraction principle resolves the contradiction by isolating the useful behavior data from the problematic identity data.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent introduces an intermediary synthetic identity generated by the generator network. This synthetic identity serves as a mediator that preserves behavior characteristics while eliminating real personal identity information. The synthetic face acts as a placeholder that maintains behavioral fidelity without compromising privacy, thus resolving the data quantity vs. accuracy contradiction.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Reliability

If anonymization is applied to preserve privacy, then identity protection is improved, but behavior information may be lost making videos uninterpretable

Engineering Contradiction:

Improveidentity protectionVSAvoidbehavior information

Core Design Contradiction:

ReliabilityVSLoss of information

Solution Approach 1:

The patent applies different processing qualities to different parts of the input data. The encoder selectively extracts behavior-related features (local quality preservation) while discarding identity-specific features. The generator then reconstructs faces with synthetic identities that maintain behavior characteristics. This local quality approach ensures identity protection while preserving behavior information integrity.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent employs feedback mechanisms through the loss function that includes behavior preservation terms. The system continuously monitors whether behavior information is maintained during the anonymization process and adjusts the generation accordingly. This feedback loop ensures that privacy protection does not come at the cost of behavior information loss.

Inventive Principle:

Principle #23Feedback

3Reliability

If a synthetic identity is generated that is completely different from all input identities, then privacy protection is improved, but the synthetic face may not resemble a real face

Engineering Contradiction:

Improveprivacy protectionVSAvoidface realism

Core Design Contradiction:

ReliabilityVSShape

Solution Approach 1:

The patent creates a composite synthetic identity by combining features from multiple input faces through the generator network. Rather than copying a single identity or creating a completely abstract representation, the generator synthesizes a new face that composite combines aesthetic features from multiple sources. This composite approach ensures both realism (by drawing from real face features) and privacy protection (by not replicating any single identity).

Inventive Principle:

Principle #40Composite materials

Data Source

PatentUS20250069366A1Generative Neural Network For Synthesis Of Faces And Behaviors

Publication Date: 2025.02.27 BLUESKEYE AI LTD

AI summary

A computing system comprising a generative neural network (100) is disclosed. The generative neural network (100) is configured to receive a plurality of input videos. Each input video comprises a face defining an identity. Each input video comprises a behaviour. Each input video comprises the same behaviour as each of the other input videos, and a different identity to each of the other input videos. The generative neural network (100) is also configured synthesize an output video from the input videos. The output video comprises a synthetic face defining a synthetic identity. The generative neural network has been trained, with a loss function, to preserve the behaviour of the input videos while generating the synthetic identity that is different from each identity of the input videos.