Hyperspherical Face Recognition Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Face recognition technologies face inaccuracies due to interference factors in images, leading to low accuracy and high misjudgment rates in identity verification.
Innovation Solution
A face recognition method and apparatus that performs feature extraction in a hyperspherical space, using both a feature extraction submodel and a prediction submodel to obtain feature vectors and values, which consider uncertainty, thereby improving accuracy by reducing reliance on feature vectors alone.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional face recognition methods are used, then the process is simple, but the accuracy is low due to interference factors in images
Solution Approach 1:
The patent transforms face feature representation from traditional Euclidean space to hyperspherical space, adding a dimensional perspective. This dimensional change allows the system to capture face features more effectively while being more robust to interference factors like lighting and pose variations, thereby improving recognition accuracy without excessively increasing complexity
Solution Approach 2:
The patent changes the parameter space by introducing hyperspherical coordinates (radius, polar angle, azimuthal angle) to represent face features instead of traditional Cartesian coordinates. This parameter transformation enables better separation of meaningful face variations from interference factors, improving measurement precision in face recognition
2Reliability
If only feature vectors are used for recognition, then the computation is fast, but misjudgment rate is high
Solution Approach 1:
The patent segments the face feature representation into multiple independent components: radius (overall face scale), polar angle (pose information), and azimuthal angle (identity information). This segmentation allows the system to process and compare different aspects of face features separately, improving reliability by considering multiple dimensions rather than relying on a single feature vector
Solution Approach 2:
The patent introduces hyperspherical space as an intermediary representation between the raw face image and the final recognition decision. This intermediary space transforms the feature representation in a way that naturally separates identity-critical features from interference factors, improving verification reliability while maintaining computational efficiency through geometric operations
3Measurement precision
If feature extraction is performed in traditional space, then the method is straightforward, but interference factors reduce accuracy
Solution Approach 1:
By mapping face features to hyperspherical space, the patent introduces a new dimensional framework where interference factors like lighting variations and pose changes are naturally separated from identity-critical features. The hyperspherical structure allows radius to capture overall scale, polar angle to represent pose, and azimuthal angle to encode identity, thereby improving measurement precision despite harmful factors
Solution Approach 2:
The patent changes the parameter representation from traditional 2D/3D Cartesian coordinates to hyperspherical parameters (radius, polar angle, azimuthal angle). This parameter transformation makes the feature representation more invariant to interference factors, as the hyperspherical coordinates naturally decouple scale, pose, and identity information, improving accuracy under varying conditions
Data Source
AI summary
A face recognition method includes: obtaining a first feature image that describes a face feature of a target face image and a first feature vector corresponding to the first feature image; obtaining a first feature value that represents a degree of difference between a face feature in the first feature image and that in the target face image; obtaining a similarity between the target face image and a template face image according to the first feature vector, the first feature value, and a second feature vector and a second feature value corresponding to a second feature image of the template face image, the second feature value describing a degree of difference between a face feature in the second feature image and that in the template face image; and determining, when the similarity is greater than a preset threshold, that the target face image matches the template face image.