Face Recognition System with Selective Feature Extraction for Masked Users
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional face recognition systems fail to properly recognize individuals wearing masks or eye bandages, as the feature values of the mouth and eyes change significantly, leading to unsuccessful identity verification and requiring users to remove these obstructions during authentication.
Innovation Solution
A face recognition system that includes a feature value selection section, allowing users to choose which facial features to prioritize for extraction, and automatically adjusts based on user conditions such as wearing a mask or eye bandage, using a database to compare selected feature values for accurate recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the face recognition system uses template matching processing requiring detection of both eyes and mouth, then the recognition accuracy is improved, but the system cannot properly recognize users wearing masks causing authentication failure
Solution Approach 1:
The patent segments the face recognition process into multiple independent feature detection modules (eye detection, mouth detection, nose detection). When a mask is detected, the system automatically switches to use only eye features for recognition, eliminating the requirement for mouth detection and enabling successful authentication of masked users while maintaining recognition accuracy through alternative feature pathways
Solution Approach 2:
The system dynamically adjusts the feature extraction requirements based on real-time detection results. A mask detection module continuously monitors the input image, and when a mask is detected, the system transitions from requiring both eye and mouth features to using only eye features. This dynamic adaptation allows the system to maintain high recognition accuracy across different user conditions (masked and unmasked)
2Reliability
If the system requires detection of mouth and both eyes for face recognition, then the reliability of authentication is improved, but the ease of operation deteriorates as users must remove masks during authentication
Solution Approach 1:
The system performs self-adjustment by automatically detecting whether the user is wearing a mask and independently switching the recognition mode accordingly. The mask detection module autonomously determines the appropriate feature set to use, and the recognition algorithm automatically adapts without requiring user intervention. This eliminates the need for users to remove masks or manually adjust settings, significantly improving ease of operation while maintaining authentication reliability
Solution Approach 2:
The system changes the recognition parameters (feature requirements) based on the detected condition. When no mask is detected, the system uses the standard parameter set requiring both eyes and mouth. When a mask is detected, the parameters are automatically changed to require only eyes for recognition. This parameter adaptation maintains reliable authentication across different conditions while allowing users to operate the system conveniently with masks worn
3Manufacturing precision
If the face recognition system uses fixed feature extraction from specific facial parts, then the manufacturing precision of the recognition algorithm is improved, but the adaptability to changing user conditions (masks, eye bandages, hairstyle changes) deteriorates
Solution Approach 1:
The patent implements a universal face recognition framework that can function with multiple different feature sets. The system includes multiple detection modules (eyes, mouth, nose) and can selectively combine them based on user conditions. This multi-functionality allows the same recognition algorithm to maintain high precision whether the user is wearing a mask, eye bandage, or has changed hairstyle, as the system can adaptively select the appropriate features to extract and use for recognition
Solution Approach 2:
The system dynamically adjusts which facial features are extracted and used for recognition based on real-time detection of user conditions. When eye bandages or hairstyle changes are detected, the system automatically switches to prioritize mouth and nose features. This dynamic feature selection maintains the precision of the recognition algorithm across various user conditions without requiring manual reconfiguration or reducing adaptability
Data Source
AI summary
A face recognition system includes: an image input section that acquires a facial image of a user; a database in which a feature value of a part of a face of a registered person is stored; a facial position detection section that detects a position of a part of a face from the facial image of the user; a feature value extraction section that extracts a feature value of the part of the face; and a recognition section that recognizes the user by comparing the feature value of the part of the face which is extracted by the feature value extraction section with the feature value of the part of the face of the registered person stored in the database, in which the user selects the part of the face from which the feature value is extracted.