Face Liveness Detection via Eye Closity and Blink Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face authentication systems are vulnerable to facial forgery, particularly through still-image based spoofing attacks, which can deceive the system by using rotated or bent photographs, and current liveness detection techniques suffer from sensitivity to noise, low accuracy, and require pre-existing models or pre-training.
Innovation Solution
The system detects facial liveness by analyzing motion in captured image frames, such as eye blinks, mouth movements, and head rotations, without the need for pre-training, using closity values and angle calculations to determine the liveness of the face, and combines multiple detection techniques for robust authentication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional face authentication techniques are used, then the authentication process is simple and fast, but the system is vulnerable to facial forgery attacks
Solution Approach 1:
The system transitions from static image analysis to dynamic video frame analysis, capturing temporal changes in facial features. By analyzing motion patterns across multiple frames (eye blinks, head movements, facial expressions), the system dynamically detects liveness without requiring complex pre-trained models, thus improving reliability while maintaining operational simplicity
Solution Approach 2:
The system performs preliminary geometric analysis by establishing reference feature points and calculating baseline geometric relationships before conducting liveness detection. This preliminary setup creates a framework for detecting anomalies in real-time without requiring complex machine learning models during the actual authentication process
2Measurement precision
If pre-trained models are used for liveness detection, then detection accuracy may be improved, but the system requires pre-existing models and pre-training processes
Solution Approach 1:
The system performs self-calibration by automatically establishing geometric relationships between facial feature points during the authentication process itself. It calculates closity values and geometric anomalies in real-time without requiring external pre-trained models, making the system easy to deploy while maintaining detection accuracy through mathematical geometric analysis
Solution Approach 2:
The system replaces complex machine learning-based liveness detection with a mathematical geometric analysis approach. By using geometric relationships between facial feature points and calculating closity values through deterministic mathematical operations, the system achieves accurate liveness detection without requiring pre-trained neural networks or complex AI models
3Object-affected harmful factors
If geometric anomaly detection is used to detect rotated or bent photographs, then forgery detection capability is improved, but the detection system becomes more complex
Solution Approach 1:
The system segments the face into multiple geometric feature points (eyes, nose, mouth, contours) and analyzes the spatial relationships between them. By dividing the facial geometry into discrete measurable points and calculating their relative positions and angles, the system detects rotations and distortions through simple geometric comparisons rather than complex holistic analysis
Solution Approach 2:
The geometric feature point analysis framework serves multiple functions simultaneously: it enables face detection, liveness detection through closity value calculation, and forgery detection through geometric anomaly analysis. This universal approach handles multiple security requirements without requiring separate complex detection systems for each function
Data Source
AI summary
Disclosed herein are techniques for face-based user authentication. In one embodiment, a method includes receiving a sequence of image frames captured of a face of a subject, and calculating, for each image frame in a set of image frames from the sequence of image frames, a closity value for the image frame based upon a plurality of angles associated with an eye in the image frame. The closity value calculated for the image frame is indicative of a measure of closeness of the eye in the image frame. The method further includes determining a number of eye blinks occurring in the set of image frames based upon the closity values calculated for the set of image frames, determining liveness of the face of the subject based upon the number of eye blinks, and enabling authentication of the subject based upon the liveness determination.