Biometric Identification Bias Correction by Appearance Category
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Biometric identification systems exhibit biases leading to identification failures and discrimination against individuals based on physical and ethnic physiological characteristics, necessitating human intervention to correct these errors.
Innovation Solution
A method to determine correctives for statistical identification variances by categorizing individuals based on physical appearance characteristics and adjusting the number of individuals for manual verification to equalize failure rates across categories, using a data processing device and convolutional neural networks for feature extraction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If biometric identification algorithms are used for automatic identification, then productivity is improved, but reliability deteriorates due to identification biases against certain physical appearance groups
Solution Approach 1:
The patent applies local quality by implementing category-specific correction factors tailored to different physical appearance groups (e.g., skin color, age, gender). Each category receives customized adjustment based on its specific identification failure patterns, rather than applying a uniform correction across all individuals. This allows the system to address reliability issues for specific groups while maintaining overall automated identification throughput.
Solution Approach 2:
The patent changes the parameter of identification thresholds and scoring criteria based on physical appearance categories. By adjusting acceptance thresholds, similarity match criteria, and confidence level requirements specific to each category, the system compensates for algorithmic biases. This parameter adjustment ensures that individuals from under-represented groups receive appropriate identification accuracy without requiring manual review of all cases.
2Reliability
If manual verification is increased to correct identification biases, then reliability is improved, but productivity deteriorates due to increased human intervention
Solution Approach 1:
The patent applies partial action by implementing selective manual verification only for specific categories where identification biases are detected, rather than requiring manual review for all individuals. The system calculates correction factors to determine the appropriate level of manual verification needed for each category, applying verification requirements proportionally to the identified bias severity. This maintains high throughput for unbiased categories while ensuring reliability for biased categories.
Solution Approach 2:
The patent implements feedback mechanisms where identification results are continuously monitored and analyzed by category. The system uses this feedback to dynamically adjust correction factors and manual verification requirements, creating a closed-loop system that optimizes the balance between automated processing and manual verification. This feedback-driven approach ensures reliability improvements without permanently reducing productivity.
3Device complexity
If uniform identification thresholds are applied to all individuals, then device complexity is reduced, but manufacturing precision deteriorates due to inability to correct category-specific biases
Solution Approach 1:
The patent applies segmentation by dividing the identification population into distinct categories based on physical appearance characteristics (skin color, age groups, gender, etc.). Each segment receives category-specific correction factors and verification requirements. This segmentation allows the system to maintain relatively simple base algorithms while adding targeted precision for each group through separate correction parameters, rather than requiring completely complex individualized algorithms for every person.
Data Source
AI summary
A method is provided for determining correctives to statistical identification variances of a biometric identification system using facial and/or pedestrian recognition. The method includes (a) defining a set of categories, each category having at least one corresponding physical appearance characteristic; (b) extracting, for each individual, at least one physical appearance characteristic from at least one image of the individual; (c) assigning, to each individual, at least one category on the basis of the extracted physical appearance characteristic; (d) distributing, for each category, each individual into two groups according to the failure or success of their identification by the system; and (e) calculating, for each category, a number of individuals to be selected in the group corresponding to the success of the identification such that the relative proportions of individuals in the group corresponding to the failure of the identification are substantially equal between all the criteria relating to said category.