Biometric Verification Probability Normalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current biometric verification systems lack a standardized method to determine the similarity between input and reference biometric identifiers across different platforms and algorithms, leading to inconsistent and unintuitive matching scores, which complicates the verification process and access control.
Innovation Solution
A computer-implemented method that calculates verification probabilities by determining matching scores between an input biometric identifier and both a cohort and a reference biometric identifier, using pre-trained distributions for genuine and imposter comparisons, allowing for normalization and efficient computation of verification probabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional biometric verification systems compare input biometric identifiers directly against reference identifiers using fixed thresholds, then the verification process is simple to implement, but the matching scores are inconsistent across different platforms and algorithms and lack standardized interpretation
Solution Approach 1:
The patent transforms matching scores from algorithm-specific values into standardized probabilities by changing the parameter representation. It uses statistical distributions (Gaussian models) to convert raw matching scores into normalized probability values between 0 and 1, enabling consistent interpretation across different biometric algorithms and platforms while maintaining computational efficiency through pre-trained distribution parameters
Solution Approach 2:
The patent introduces statistical distributions as an intermediary layer between the biometric matching process and the verification decision. By modeling matching scores through Gaussian distributions with pre-determined mean and standard deviation parameters, the system mediates between different algorithms and provides a unified probability-based verification framework that is both accurate and computationally efficient
2Reliability
If biometric verification uses comprehensive comparison methods across multiple identifiers, then the verification accuracy improves, but the computational time and processing speed increase
Solution Approach 1:
The patent performs preliminary actions by pre-training and storing statistical distribution parameters (mean and standard deviation) for both genuine and imposter matching scores during a setup phase. During actual verification, the system only needs to compute a simple probability using these pre-computed parameters, avoiding the need for time-consuming comprehensive comparisons while maintaining high verification accuracy through the statistically optimized probability calculation
3Measurement precision
If the system uses complex algorithms to determine biometric similarity, then the matching precision improves, but the hardware requirements and system cost increase
Solution Approach 1:
The patent replaces complex mechanical or computational biometric verification systems with a statistically simplified approach. By substituting elaborate real-time comparison algorithms with pre-computed Gaussian distribution models and simple probability calculations, the system achieves high matching precision using inexpensive hardware that can perform basic arithmetic operations, eliminating the need for expensive specialized processing units
Data Source
AI summary
A method of verifying an input biometric identifier against a reference biometric identifier is disclosed in this specification. The method comprises evaluating the input biometric identifier relative to a group (the ‘cohort’) to improve verification accuracy. Up to three matching scores are used to determine a verification probability for the input biometric identifier. The three matching scores measure the similarity of the input biometric identifier to the biometric identifiers of the cohort, the similarity of the reference biometric identifier to the biometric identifiers of the cohort and the similarity of the input biometric identifier to the reference biometric identifier.