Multi-Dictionary Registration for Face Recognition Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing personal authentication methods face challenges in efficiently and accurately identifying individuals, particularly when attribute determination errors occur, leading to incorrect dictionary registration and increased identification processing time and errors, especially with androgynous faces.
Innovation Solution
An information processing apparatus and method that includes an obtaining unit, detection unit, attribute determination unit, registration unit, and adding unit to register image data in multiple dictionaries and add link information, allowing for correction of attribute determination errors and improved identification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If image data is registered in only one dictionary based on attribute determination, then identification processing speed is improved, but identification accuracy deteriorates when attribute determination errors occur
Solution Approach 1:
The patent divides the single dictionary into multiple dictionaries based on attributes (e.g., male/female, young/old). Image data is registered in multiple dictionaries corresponding to different attribute categories. This segmentation allows the system to quickly narrow down search scope using attribute matching while maintaining accuracy by having redundant storage across multiple attribute-based dictionaries.
Solution Approach 2:
The patent performs attribute determination and pre-registers image data in multiple appropriate dictionaries before actual identification occurs. This preliminary action ensures that when identification is needed, the system only needs to search within pre-filtered dictionaries matching the input image's attributes, significantly reducing search time while maintaining accuracy through redundant storage.
2Reliability
If image data is registered in multiple dictionaries, then identification accuracy is improved, but identification processing load increases
Solution Approach 1:
The patent applies different storage strategies to different dictionaries based on their specific attributes. Each dictionary is optimized for its particular attribute category (e.g., gender-based, age-based), allowing the system to register image data selectively in multiple dictionaries only when attribute determination confidence is sufficient, thereby managing processing load while maintaining accuracy.
3Speed
If attribute determination is performed to narrow down dictionaries, then identification speed is improved, but identification performance deteriorates when attribute determination is unreliable
Solution Approach 1:
The patent prepares multiple dictionaries in advance with different attribute classifications, creating a cushion against attribute determination errors. When attribute determination is performed, the system has pre-prepared dictionaries to fall back on, ensuring that even if attribute determination is unreliable, the identification can still proceed using alternative attribute-based dictionaries, thus maintaining both speed and reliability.
Data Source
AI summary
An information processing apparatus comprising: an obtaining unit configured to obtain image data; a detection unit configured to detect an object from the image data; an attribute determination unit configured to determine an attribute indicating a characteristic of the object detected by the detection unit; a registration unit configured to register the image data in at least one of a plurality of dictionaries based on the attribute determined by the attribute determination unit; and an adding unit configured to add, when the image data is registered in not less than two dictionaries, link information concerning the image data registered in the other dictionary to the image data registered in one dictionary.