Biometric Model Adaptation Using Synthetic Environment Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing biometric authentication systems face performance deterioration when the training environment differs from the actual environment due to the need for collecting sensitive biometric information, which raises privacy concerns and user reluctance.
Innovation Solution
A machine learning system that utilizes biometric attribute features and environment attributes to update a distribution parameter, generating combined biometric information for training without collecting actual environment data, using a client-server architecture to adapt the model to the actual environment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If biometric information is collected from the actual environment for training, then authentication accuracy is improved, but privacy protection is worsened and user reluctance increases
Solution Approach 1:
The patent creates synthetic biometric data by combining extracted attribute features with generated identification features. This copying approach allows the system to train using artificial representations of biometric data that reflect actual environment conditions without collecting real user biometric information, thus maintaining authentication accuracy while protecting privacy.
Solution Approach 2:
The patent introduces an intermediary processing layer that extracts attribute features from actual environment biometric data and uses these to generate synthetic training data. This intermediary mechanism enables the system to learn from actual environment characteristics without directly handling or storing sensitive user biometric information, resolving the privacy accuracy tradeoff.
2Adaptability or versatility
If biometric information is collected from the actual environment, then model performance in actual environment is improved, but user cooperation is worsened
Solution Approach 1:
The system creates synthetic biometric samples by combining extracted attribute features with generated identification features. Users do not need to provide actual biometric information for training, as the system generates synthetic data that captures actual environment characteristics. This eliminates user reluctance while maintaining model adaptability to real-world conditions.
Solution Approach 2:
The system performs self-service by automatically extracting attribute features from actual environment data and generating synthetic training data without requiring user participation. The machine learning device autonomously adapts to actual environment conditions through this self-service process, eliminating the need for user cooperation while maintaining environmental adaptability.
3Reliability
If training data is collected from actual environment, then authentication accuracy under various conditions is improved, but data processing complexity is worsened
Solution Approach 1:
The patent segments the biometric data processing into distinct components: extracting attribute features from actual environment data, generating identification features, and combining them to create synthetic training data. This segmentation allows the system to handle complex environmental variations systematically through modular processing steps, reducing overall complexity while improving authentication accuracy under various conditions.
Data Source
AI summary
A machine learning device: updates a distribution parameter based on the actual environment attribute feature corresponding to biometric information on a target user; extracts an attribute feature from the updated distribution parameter; generate combined biometric information based on the extracted attribute feature and a training identification feature extracted from training biometric information; and updates a machine learning model based on the combined biometric information and the training biometric information, the attribute feature is a feature that has an influence on an authentication accuracy of biometric authentication and has a low correlation with an identification feature, and the identification feature is used for collation in the biometric authentication.