User Verification Using Generalized Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing user verification systems face challenges in accurately distinguishing between enrolled users and generalized users, leading to inefficient verification processes and increased false negatives or false positives.
Innovation Solution
A user verification method and apparatus utilizing a neural network-based system that generates test feature vectors and employs comparators to determine similarity parameters, adjusting verification conditions based on both enrolled and generalized user models, thereby enhancing reliability and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional user verification systems compare test feature vectors only with enrolled user models, then the verification process is simple, but the accuracy decreases leading to increased false negatives and false positives
Solution Approach 1:
The verification system is segmented into multiple comparison paths: one path compares test feature vectors with enrolled user models for authenticated users, while another path compares with generalized user models for unauthenticated users. This segmentation allows the system to handle different user types with appropriate verification strategies, improving overall accuracy without requiring a complete redesign of the verification architecture.
Solution Approach 2:
Generalized user models serve as an intermediary component between the test feature vectors and the final verification decision. These models represent aggregate characteristics of unauthenticated users and act as a reference point for comparison. By introducing this intermediary, the system can distinguish between genuine users and impostors more effectively, reducing false positives and false negatives.
2Reliability
If the verification system uses fixed verification conditions for all users, then the system is easy to operate, but the reliability decreases due to inability to adapt to different user types
Solution Approach 1:
The verification conditions are made dynamic by automatically selecting different comparison strategies based on user authentication status. For authenticated users, the system dynamically switches to comparing against their specific enrolled models with appropriate thresholds. For unauthenticated users, it dynamically switches to comparing against generalized user models. This dynamic adaptation improves reliability without requiring manual configuration of multiple verification policies.
Solution Approach 2:
The system incorporates feedback mechanisms where the verification outcome and user authentication status inform the selection of verification conditions. The system learns from verification results and adjusts the weighting between enrolled user model comparisons and generalized user model comparisons, thereby improving reliability over time while maintaining ease of operation through automatic adaptation.
3Measurement precision
If the system performs comprehensive comparisons with multiple user models, then the verification accuracy improves, but the processing time increases
Solution Approach 1:
The system performs partial comparisons by selectively engaging different verification paths based on user authentication status. For unauthenticated users, only comparisons with generalized user models are performed rather than exhaustive comparisons with all possible user models. This partial action approach maintains verification accuracy for the relevant user type while significantly reducing processing time by avoiding unnecessary comparisons.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A user verification method and apparatus using a generalized user model is disclosed, where the user verification method includes generating a feature vector corresponding to a user based on input data corresponding to the user, determining a first parameter indicating a similarity between the feature vector and an enrolled feature vector enrolled for user verification, determining a second parameter indicating a similarity between the feature vector and a user model corresponding to generalized users, and verifying the user based on the first parameter and the second parameter.