Facial Landmark Sparse Code Authentication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current user authentication methods, particularly for secure locations, face challenges in providing robust and efficient verification processes that can accurately identify users based on facial recognition, especially in environments where security is stringent and traditional methods may fail due to variations in image quality or distortion.
Innovation Solution
A user authentication method that extracts facial landmarks from input images, generates sparse codes using feature vectors based on learning data, and determines user authentication by calculating similarity between these sparse codes, employing techniques like cosine or chi-square distance measurements to ensure accurate identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional facial recognition methods are used, then the authentication process is simple, but the accuracy decreases due to variations in image quality or distortion
Solution Approach 1:
The patent divides the face into multiple local regions (eyes, eyebrows, nose, mouth, chin) and extracts features from each region separately. This segmentation allows the system to focus on distinctive local characteristics rather than relying on the entire face image, thereby improving authentication accuracy while managing complexity through modular processing
Solution Approach 2:
The patent transforms the authentication approach from direct image comparison to sparse code representation in a different mathematical dimension. By converting facial features into sparse codes through dictionary learning, the system operates in a transformed feature space that is more robust to image quality variations and distortions
2Measurement precision
If sparse codes are generated using feature vectors based on learning data, then the authentication accuracy improves, but the processing time increases
Solution Approach 1:
The patent performs dictionary learning and feature extraction in advance during an enrollment phase. By pre-processing and storing reference sparse codes for each user, the system reduces the computational burden during actual authentication, thereby decreasing processing time while maintaining high accuracy through pre-computed feature representations
Solution Approach 2:
The patent creates sparse code representations as compressed copies of facial features. Instead of storing or processing entire high-resolution images, the system uses compact sparse codes that capture essential facial characteristics, significantly reducing processing time and computational resources while preserving authentication accuracy
3Reliability
If multiple facial landmarks are extracted and processed, then the robustness of authentication improves, but the computational complexity increases
Solution Approach 1:
The patent extracts only the most discriminative facial landmarks (key points of eyes, eyebrows, nose, mouth, chin) rather than processing the entire face image. This selective extraction of critical features maintains authentication robustness by focusing on distinctive characteristics while reducing computational complexity through targeted processing
Solution Approach 2:
The patent applies different processing strategies to different facial regions based on their local characteristics. By treating each facial landmark region with appropriate feature extraction methods tailored to its specific properties, the system achieves robust authentication while optimizing computational resources for each local area rather than applying uniform complex processing
Data Source
AI summary
A user authentication method includes extracting a facial landmark from each of a first input image and a second input image; generating a first sparse code of the facial landmark extracted from the first input image and a second sparse code of the facial landmark extracted from the second input image; and determining whether a user is to be authenticated based on the first sparse code of the first input image and the second sparse code of the second input image.