Multi-Task Biometric Neural Network Architecture
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing biometric systems face challenges in accurately identifying users and detecting spoofing attempts, particularly due to redundant tasks in facial biometric algorithms and limited training datasets.
Innovation Solution
A multi-task neural network architecture is introduced, featuring a common feature extraction neural network and multiple biometric analysis task neural networks, which share latent variables and utilize expert pooling to enhance training and reduce computational resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple separate neural networks are used for different biometric tasks, then task-specific accuracy is improved, but device complexity and computational burden increase
Solution Approach 1:
The patent combines multiple separate biometric analysis neural networks into a single multi-task neural network that shares a common feature extraction backbone. This merging reduces device complexity while maintaining task-specific accuracy through shared latent variables and task-specific output layers.
Solution Approach 2:
The common feature extraction neural network serves as a universal backbone that processes inputs for multiple different biometric tasks simultaneously. This multi-functional design allows a single network to perform identification, authentication, and other biometric analyses without requiring separate dedicated networks for each task.
2Measurement precision
If more training data is collected for comprehensive biometric analysis, then identification accuracy is improved, but data processing time and computational resources increase
Solution Approach 1:
The patent merges multiple task-specific training processes into a single unified training procedure that optimizes all biometric tasks simultaneously. By sharing the common feature extraction network and using joint loss functions, the system achieves comprehensive accuracy without requiring separate extensive training campaigns for each task.
Solution Approach 2:
The multi-task neural network enables continuous learning across all biometric tasks through a unified training framework. The shared backbone network continuously refines feature extraction capabilities that benefit all tasks, maintaining continuous improvement in accuracy without discrete training interruptions for each individual task.
3Reliability
If comprehensive feature extraction is performed for all biometric tasks, then reliability is improved, but computational power requirements increase
Solution Approach 1:
The patent merges redundant feature extraction operations across multiple tasks into a single shared common backbone network. This consolidation maintains comprehensive feature analysis for reliable authentication while eliminating redundant computational operations that would otherwise be performed separately by multiple independent networks.
Solution Approach 2:
The common feature extraction neural network is designed as a universal processor that efficiently extracts features useful for multiple biometric tasks simultaneously. This multi-functional extraction reduces the total computational power required compared to having separate specialized networks, as the shared backbone learns to extract versatile features that serve multiple purposes.
Data Source
AI summary
A deep neural network can provide output from a selected biometric analysis task that is one of a plurality of biometric analysis tasks based on an image. The selected biometric analysis task can be performed in a deep neural network that includes a common feature extraction neural network, a plurality of biometric task-specific neural networks, a plurality of segmentation mask neural networks and an expert pooling neural network that perform the plurality of biometric analysis tasks by inputting the image to the common feature extraction network to determine latent variables. The latent variables can be input to the plurality of biometric task-specific neural networks to determine a plurality of biometric analysis task outputs. The latent variables can be input to a segmentation neural network to determine a facial feature segmentation output. The facial feature segmentation output can be output to a plurality of segmentation mask neural networks.