Fingerprint Block Classification Without Gabor Filtering Overhead
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for obtaining biometric features, such as fingerprint minutiae, require sophisticated neural networks and additional filtering steps like Gabor filtering, which are computationally expensive and resource-intensive.
Innovation Solution
A computer-implemented method that processes images using a neural network to identify biometric features by separating them into blocks, reducing the need for orientation maps or segmentation masks, and providing a binary decision on feature presence, with optional additional information like quality, position, and orientation, using a pre-trained neural network on mobile devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sophisticated neural networks and additional filtering steps like Gabor filtering are used to identify biometric features, then measurement precision is improved, but use of energy by moving object and device complexity increase
Solution Approach 1:
The patent divides the fingerprint image into multiple blocks of identical size, where each block is independently processed by the neural network. This segmentation allows the system to process only relevant portions of the image, reducing overall computational load while maintaining identification accuracy through localized feature analysis.
Solution Approach 2:
The patent extracts and removes the need for additional filtering steps like Gabor filtering and orientation maps from the processing pipeline. By using a neural network that can directly process fingerprint images and identify minutiae without these intermediate processing stages, the system reduces computational complexity and energy consumption while preserving measurement precision.
2Measurement precision
If sophisticated neural networks and additional filtering steps are used to identify biometric features, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent merges the functions of multiple separate processing components (Gabor filtering, orientation map generation, segmentation mask creation, and neural network classification) into a single integrated neural network model. This consolidation eliminates the need for multiple separate algorithms and data structures, thereby reducing device complexity while maintaining or improving measurement precision.
Solution Approach 2:
The patent extracts and eliminates unnecessary intermediate processing steps from the traditional fingerprint analysis pipeline. By removing the need for Gabor filtering, orientation maps, and segmentation masks, the system simplifies its architecture while still achieving accurate biometric feature identification through the neural network's direct processing capability.
3Measurement precision
If orientation maps or segmentation masks that have to be trained are used, then measurement precision is improved, but use of energy by moving object increases
Solution Approach 1:
The patent extracts and removes the training requirements for orientation maps and segmentation masks from the system. The neural network is designed to process fingerprint images directly without requiring these pre-computed maps, thereby eliminating the significant computational energy that would be required for their training and generation while maintaining identification accuracy.
Data Source
AI summary
A computer-implemented method for obtaining information on biometric features of a fingerprint of a user, the method comprising obtaining an image comprising an object carrying one or more biometric features of a user, the image having a size of N×M pixels, processing the image and determining information indicative at least of a presence of a biometric feature in the block and a kind of the biometric feature to obtain a result, and outputting the result identifying the information wherein the result has a form of n×m blocks of identical size and the result associates information indicative at least of a presence of a biometric feature and a kind of the biometric feature with each block, wherein N is an integer multiple of n and M is an integer multiple of m.