Face Recognition Neural Network Using Multi-Feature Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face recognition methods using Convolutional Neural Networks (CNNs) face performance issues under special conditions like low light and generate redundant features, reducing efficiency and increasing costs due to the use of multiple CNNs and simple feature vectors that do not consider complementary features.
Innovation Solution
A method involving a neural network with a first sub-neural network generating deep feature vectors from preliminary feature vectors, including color-based, texture-based, and gradient-based features, and a second sub-neural network for pose determination, using error backpropagation and feature fusion to improve recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple CNNs are used to process different sub-images and fuse feature vectors, then face recognition performance is improved, but device complexity and cost increase
Solution Approach 1:
The patent divides the feature extraction process into multiple stages by using multiple CNNs to process different sub-images (eye region, nose region, mouth region, etc.) separately, then fuses the extracted feature vectors. This segmentation allows each CNN to specialize in specific facial regions, improving overall recognition performance while managing complexity through modular architecture
Solution Approach 2:
The patent merges the output of multiple CNNs by fusing their extracted feature vectors at the feature layer. This combining approach integrates information from different facial regions and CNN models, achieving improved recognition accuracy while sharing computational resources efficiently
2Speed
If simple feature vectors are used as input, then processing speed is improved, but recognition accuracy deteriorates due to lack of complementary features
Solution Approach 1:
The patent creates composite feature vectors by combining multiple types of simple feature vectors (color-based, texture-based, gradient-based, etc.) that capture different aspects of facial characteristics. This composite approach enriches the feature representation without requiring a single complex CNN, maintaining processing speed while improving recognition accuracy through complementary information
3Device complexity
If only greyscale or RGB images are used as input, then processing is simplified, but performance under special conditions (e.g., low light) deteriorates
Solution Approach 1:
The patent designs a multi-functional feature extraction system that can process multiple types of input images (greyscale, RGB, and other specialized images) through a unified architecture of multiple CNNs. Each CNN is trained to handle specific image types and lighting conditions, making the system universally applicable to various imaging scenarios including low light conditions without requiring separate specialized systems
Data Source
AI summary
Systems and methods for face recognition are provided. The systems may perform the methods to obtain a neural network comprising a first sub-neural network and a second sub-neural network; generate a plurality of preliminary feature vectors based on an image associated with a human face, the plurality of preliminary feature vectors comprising a color-based feature vector; obtain at least one input feature vector based on the plurality of preliminary feature vectors; generate a deep feature vector based on the at least one input feature vector using the first sub-neural network; and recognize the human face based on the deep feature vector.