Facial Mesh Liveness Detection for Mobile Authentication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Biometric face-based authentication systems are vulnerable to facial forgery and face liveness detection on mobile devices is computationally challenging due to the lack of integrated GPUs, and cloud-based solutions require significant bandwidth and time for high-resolution video processing.
Innovation Solution
A method and system using facial mesh analysis on portable devices to detect liveness by generating initial and difference facial meshes based on facial landmarks, with prompts to evoke emotional states, and evaluating these meshes using a machine learning algorithm, either locally or remotely.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If cloud-based processing is used for liveness detection, then processing power and accuracy are improved, but bandwidth consumption and transmission time increase significantly
Solution Approach 1:
The patent segments the liveness detection system into two parts: lightweight preprocessing and feature extraction performed locally on the mobile device, and heavier machine learning model execution performed remotely on a server. This segmentation allows the mobile device to handle initial image capture, facial mesh generation, and prompt display while transmitting only essential data (facial mesh features and predicted emotional states) to the server, dramatically reducing bandwidth consumption while maintaining processing power for complex analysis.
Solution Approach 2:
The patent introduces an intermediary processing layer that generates facial meshes and extracts key features from raw images before transmission. This intermediary step transforms high-volume image data into compact feature representations (facial mesh coordinates, edge distances, and emotional state predictions), serving as a mediator between the mobile device camera and the cloud-based machine learning model, thereby reducing the data transmission burden while preserving essential information for accurate liveness detection.
2Measurement precision
If deep neural networks are used for facial expression analysis, then accuracy is improved, but computational complexity and device requirements increase
Solution Approach 1:
The patent segments the computational workload by dividing facial expression analysis into two stages: a lightweight initial assessment performed on the mobile device using simple image processing to generate facial meshes and detect basic expressions, and a more sophisticated deep neural network analysis performed remotely on the server. This segmentation enables accurate expression analysis while keeping mobile device computational requirements minimal.
Solution Approach 2:
The patent creates simplified copies or representations of facial expressions through facial mesh models that capture essential geometric features without requiring full-resolution image processing. Instead of analyzing raw high-resolution images directly with complex neural networks, the system generates mesh-based representations that preserve expression characteristics while reducing computational complexity, allowing accurate expression detection with lighter processing requirements.
3Measurement precision
If high-resolution video is captured at 25 fps, then image quality is improved, but bandwidth requirements and processing time increase
Solution Approach 1:
The patent extracts only the essential features from high-resolution video frames by generating facial meshes that capture key geometric relationships between facial landmarks. Instead of transmitting or processing complete high-resolution video streams, the system extracts and processes only the relevant facial structure information (mesh coordinates, edge distances, and landmark positions), dramatically reducing bandwidth requirements while maintaining sufficient image quality for accurate liveness detection and expression analysis.
Data Source
AI summary
There is provided a computer implemented method, computer readable medium and system for verifying liveness of a subject depicted in a plurality of acquired images. An initial facial mesh is generated from multiple images in acquired images of the face of the subject. Further images are acquired subsequent to communicating to the subject a prompt indicative of an emotional state. A difference facial mesh is generated from facial landmarks extracted from a further image; and the change of some predetermined edges relative to corresponding edges of the initial facial mesh determined. A model trained by a machine learning algorithm evaluates whether the difference facial mesh corresponds to an expected emotional state for that subject following the prompt. The above steps are repeated and if a predetermined number of a difference meshes for a subject correspond to expected emotional states the subject is verified as live.