Biometric Pattern Matching Threshold Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Biometric pattern matching systems face challenges in identifying a single conforming match from multiple matching patterns, leading to potential false acceptance and erroneous access control decisions, especially when multiple patterns from different sources match a pattern-under-test.
Innovation Solution
A method that iteratively increases the match threshold until a single pattern is identified as the conforming match, using a matching engine to evaluate and rank patterns based on a conformance standard, ensuring accurate identification even when multiple patterns initially match the pattern-under-test.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If the matching process stops upon a first match condition, then processing speed is improved, but identification accuracy deteriorates due to potential false acceptance when multiple patterns match
Solution Approach 1:
The matching process dynamically adjusts the match threshold during processing. It begins with a lower threshold to quickly identify candidate matches, then progressively increases the threshold to eliminate false positives. This dynamic adjustment allows the system to balance speed and accuracy by adapting the stringency of matching criteria based on the number of matches found.
Solution Approach 2:
The system performs a preliminary matching pass with a lower threshold to quickly identify all potential matches, including false positives. This preliminary action separates the fast identification phase from the subsequent verification phase, allowing speed in the initial scan while ensuring accuracy through later threshold-based filtering.
2Reliability
If the match threshold is increased to reduce false acceptance, then identification accuracy is improved, but processing time increases due to additional cycles required
Solution Approach 1:
The match threshold is dynamically adjusted across multiple processing cycles rather than being statically high from the beginning. The system starts with a lower threshold to quickly narrow down candidates, then incrementally increases the threshold in subsequent cycles only for the remaining candidates. This dynamic approach reduces the overall processing time compared to applying a high threshold to all patterns from the start.
Solution Approach 2:
The matching process is segmented into multiple cycles, each with progressively higher thresholds. Instead of performing one lengthy comparison with a high threshold, the system divides the work into stages: first cycle identifies candidates with lower threshold, second cycle refines the list with medium threshold, and subsequent cycles achieve final accuracy with high threshold. This segmentation reduces total processing time by eliminating early comparisons that would fail anyway.
3Adaptability or versatility
If multiple patterns from different sources are stored in the pattern storage, then system versatility is improved, but matching reliability deteriorates due to increased false acceptance risk
Solution Approach 1:
The system applies different match thresholds to different patterns or pattern sources based on their reliability characteristics. Patterns from more reliable sources or those with better quality metrics can be matched with lower thresholds, while patterns from less reliable sources require higher thresholds. This local quality adjustment maintains versatility across multiple pattern sources while preserving overall matching reliability.
Data Source
AI summary
A system and method for providing a matching process that determines a conforming pattern match to a pattern-under-test from a set of matching patterns in a pattern storage. A matching process tests every template against the pattern-under-test and resolves a conforming match condition when multiple matches to the pattern-under-test are found. In some embodiments, a matcher engine using variable match thresholds may be used to differentiate among matching templates to identify a conforming-matching template with respect to the pattern-under-test.