Facial Descriptor Weighting for Spoof-Resistant User Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electronic devices face challenges in accurately identifying users due to false detection by mimicking facial expressions or using masks, leading to security issues and unintended unlocking, which compromises user privacy.
Innovation Solution
The method involves capturing multiple facial descriptors from image frames, generating first facial descriptor coordinates, and comparing them with stored second coordinates, assigning weights based on descriptor position and motion, and resizing these coordinates to determine a second distance threshold for accurate user identification, considering facial behavioral traits and environmental factors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If facial recognition is performed by comparing real-time image with pre-stored image, then user identification can be provided, but false detection occurs where another user can mimic the original user by imitating facial expressions or using masks
Solution Approach 1:
The patent segments the facial recognition process into multiple independent analysis dimensions: (1) extracting multiple facial descriptors (eye shape, nose shape, mouth shape, etc.) from the face image, (2) determining positional relationships between these descriptors, (3) analyzing facial expressions and movements, and (4) comparing all these segmented features against stored biometric data. This segmentation allows the system to detect spoofing attempts by examining individual descriptor characteristics rather than relying on a single holistic comparison.
Solution Approach 2:
The patent applies local quality by assigning different weights and analysis depths to different facial descriptors based on their spoofing resistance characteristics. For example, certain facial descriptors like eye shape and nose structure are inherently more difficult to replicate with masks compared to facial expressions. The system analyzes each descriptor's local characteristics (shape, position, size) with appropriate precision and combines them to form the overall authentication decision, thereby improving reliability without uniformly increasing complexity across all features.
2Ease of operation
If facial recognition is performed without considering facial behavioral traits, then unlocking process is simplified, but unintended unlocking occurs when the original user is sleeping or inattentive
Solution Approach 1:
The patent introduces dynamics by analyzing temporal characteristics of facial descriptors - specifically detecting facial expressions and their movements over time. The system monitors whether facial features exhibit natural movement patterns consistent with an awake, attentive user. This dynamic analysis adds a time-based dimension to the static facial recognition process, enabling the system to distinguish between genuine users who naturally move their facial features and spoofing attempts or unconscious users who display static or unnatural facial patterns.
3Measurement precision
If multiple facial descriptors are extracted and weighted for identification, then user identification accuracy is improved, but device complexity increases
Solution Approach 1:
The patent implements partial action by selectively extracting and analyzing only the most discriminative facial descriptors rather than processing every possible facial feature. The system identifies a subset of key descriptors (such as eye shape, nose shape, mouth shape) that provide sufficient differentiation between users and assigns weights to these partial features. This approach achieves high identification accuracy by focusing computational resources on the most informative descriptors, avoiding the excessive complexity that would result from analyzing all possible facial characteristics in equal detail.
Data Source
AI summary
A method for identifying a user of an electronic device, includes: capturing at least one image frame of a portion of the user's face; extracting facial descriptors from the at least one image frame of the portion of the user's face; generating first facial descriptor coordinates by using the facial descriptors; determining a first distance between the first facial descriptor coordinates and second facial descriptor coordinates; resizing the first facial descriptor coordinates at least one of radially and angularly based on the first distance between the first facial descriptor coordinates and the second facial descriptor coordinates, and a weight associated with the facial descriptors used to generate the first facial descriptor coordinates.