Region-Weighted Eye Feature Matching for Adaptive Authentication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ensemble estimation methods for biometric authentication, such as iris recognition, struggle to effectively combine multiple estimators to enhance accuracy and reliability, particularly in varying eye conditions.
Innovation Solution
An information processing device and method that extracts feature vectors from multiple regions of an eye, calculates similarity weights, and combines scores to enhance authentication accuracy by leveraging both iris and periocular features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple estimators are used in ensemble estimation, then authentication accuracy is improved, but device complexity increases
Solution Approach 1:
The patent divides the eye region into multiple sub-regions (iris, periocular, eyelid areas) and creates separate estimators for each region. Each estimator processes feature vectors from its specific region, allowing the system to leverage multiple estimators for improved accuracy while maintaining manageable complexity through functional segmentation
Solution Approach 2:
The patent combines the estimation results from multiple region-specific estimators using weighted averaging to produce a final authentication decision. The similarity scores from iris, periocular, and other region estimators are merged with dynamically determined weights, achieving ensemble estimation benefits while integrating results through a unified decision mechanism
2Adaptability or versatility
If dynamic weighting of region scores is implemented, then adaptability to varying eye conditions is improved, but calculation complexity increases
Solution Approach 1:
The patent implements dynamic weighting where the weight assigned to each region's similarity score changes based on the actual eye image characteristics. For example, when the iris region is clearly visible, the iris estimator receives higher weight; when occluded, other regions are weighted more heavily. This dynamic adaptation allows the system to respond to varying eye conditions without fixed, rigid weighting schemes
Solution Approach 2:
The patent uses pre-stored reference feature vectors for each region that are copied and compared against feature vectors extracted from the input eye image. This copying approach enables efficient similarity calculation by reusing reference data across different estimation processes, reducing the need for complex real-time computations while maintaining adaptability
Data Source
AI summary
A feature vector of each of multiple regions, cut out from a region of an eye of a target included in an acquired image, is extracted. The weight of similarity is identified for each of the multiple regions that is calculated based on the feature vector of each of the multiple regions and a feature vector relating to each of corresponding regions that is pre-stored for the target. A similarity between feature vectors of the eye of the target included in the acquired image and pre-stored feature vectors of the eye of the target is calculated by using the feature vector of each of the multiple regions and the weights identified for those feature vectors.