Fingerprint Feature Extraction Neural Network Incremental Learning

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing fingerprint recognition technologies face challenges in gradually improving performance while minimizing resource consumption, especially when new data is collected and the model needs to be retrained, and when online learning is not suitable for real-time data handling.

Innovation Solution

A fingerprint recognition technology that continuously reflects newly input fingerprint features in an existing registered fingerprint, allowing for gradual performance improvement while minimizing resource consumption, by using a neural network for fingerprint feature extraction and modifying registered fingerprint features based on input fingerprint similarities.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If offline learning is used to train the neural network model with all collected data, then the model can be trained comprehensively, but the entire model needs to be retrained when new data is collected, consuming significant time and resources

Engineering Contradiction:

Improvefingerprint recognition accuracyVSAvoidmodel retraining time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent divides the model training process into two segments: offline learning for comprehensive model training and incremental learning for continuous improvement with new data. This segmentation allows the system to avoid complete retraining while maintaining recognition accuracy.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary comprehensive training through offline learning before deploying the model. This preliminary action establishes a solid foundation that enables subsequent incremental learning without requiring full retraining.

Inventive Principle:

Principle #10Preliminary action

2Adaptability or versatility

If online learning is used to continuously train the model with real-time data, then the model can be updated continuously, but the system performance may deteriorate if the real-time data is damaged or useless

Engineering Contradiction:

Improvecontinuous model update capabilityVSAvoidsystem performance stability

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The patent implements a feedback mechanism where the system evaluates the quality and usefulness of real-time data before incorporating it into the model. This feedback loop prevents performance deterioration by filtering out damaged or useless data while maintaining continuous update capability.

Inventive Principle:

Principle #23Feedback

3Measurement precision

If the entire model is retrained with new data, then the recognition accuracy can be improved, but the resource consumption increases significantly

Engineering Contradiction:

Improvefingerprint recognition accuracyVSAvoidcomputational resource consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent applies partial action by updating only the necessary portions of the model using incremental learning rather than retraining the entire model. This approach achieves accuracy improvement while significantly reducing computational resource consumption.

Inventive Principle:

Principle #16Partial or excessive action

4Reliability

If online learning is continuously monitored to detect performance deterioration, then the system reliability can be maintained, but the time and resources required for monitoring and stopping learning increase

Engineering Contradiction:

Improvesystem performance monitoringVSAvoidmonitoring and learning interruption time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent implements beforehand cushioning by establishing data quality evaluation mechanisms before incremental learning occurs. This preventive approach cushioning against potential performance deterioration without requiring continuous monitoring and interruption.

Inventive Principle:

Principle #11Beforehand cushioning (Prior cushioning)

Data Source

PatentUS20250182523A1Fingerprint feature extraction device, storage medium storing instructions to perform fingerprint feature extraction method, and method for training neural network to extract feature information of fingerprint

Publication Date: 2025.06.05 SUPREMA INC

AI summary

There is provided a fingerprint feature extraction device including a fingerprint contact surface. The device comprises a fingerprint input sensor configured to detect a fingerprint of a target in contact with the fingerprint contact surface; a memory configured to store instructions for outputting a fingerprint feature extraction result with respect to the fingerprint input from the fingerprint input sensor using a pre-trained neural network; and a processor configured to execute the instructions, wherein the instructions, when executed by the processor, cause the processor to modify feature information of a fingerprint registered in the fingerprint feature extraction device on the basis of feature information of the input fingerprint of the target, pair the modified feature information of the registered fingerprint with the feature information of the input fingerprint, and input results of pairing to the neural network to output the fingerprint feature extraction result.