Fingerprint Anti-Spoofing Network Training Across Multiple Sensors

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing fingerprint recognition technologies face challenges in achieving high recognition accuracy due to environmental factors and the need for separate neural network models for different sensor types, limiting their applicability and resource efficiency.

Innovation Solution

A method and apparatus for training a fingerprint anti-counterfeiting neural network that uses a fusion subnetwork to combine feature description matrices from multiple sensor types, and a classification subnetwork to recognize real and prosthetic fingerprints, incorporating a preprocessing subnetwork to convert raw data into feature description matrices and iteratively train the network using a combined loss function.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If separate neural network models are trained for different sensor types, then recognition accuracy for each sensor type is improved, but device complexity and resource consumption increase

Engineering Contradiction:
Improverecognition accuracyVSAvoidmodel training complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent trains a single neural network model using fingerprint data from multiple sensor types (capacitive, optical, ultrasonic) to achieve universal recognition capability. The model learns to handle different sensor characteristics through multi-source data training, eliminating the need for separate models for each sensor type while maintaining high recognition accuracy across all sensor types.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent merges multiple sensor types into a unified training framework by collecting and processing fingerprint data from capacitive, optical, and ultrasonic sensors together. The neural network integrates features from different sensor modalities through a common architecture, reducing overall system complexity while preserving the advantages of each sensor type.

Inventive Principle:
Principle #5Merging (Combining)

2Productivity

If neural network models are trained using fingerprint data collected by a specific sensor type, then training efficiency is improved, but the model fails to recognize fingerprints from other sensor types

Engineering Contradiction:
Improvetraining efficiencyVSAvoidapplication range
Core Design Contradiction:
ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent creates a universal neural network model that can process fingerprint data from multiple sensor types through a single training process. The model is designed with architecture and training protocols that enable it to generalize across different sensor modalities, allowing one model to serve multiple sensor types without requiring separate training for each.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent employs parameter adjustment techniques during training to adapt the neural network to different sensor characteristics. By dynamically adjusting training parameters and using multi-source data, the model learns to compensate for variations between sensor types, enabling it to maintain high accuracy across capacitive, optical, and ultrasonic sensors with a single trained model.

Inventive Principle:
Principle #35Parameter changes

3Speed

If neural network models are trained directly using fingerprint data collected by a sensor, then training speed is improved, but recognition accuracy decreases due to environmental factors

Engineering Contradiction:
Improvetraining speedVSAvoidrecognition accuracy
Core Design Contradiction:
SpeedVSMeasurement precision

Solution Approach 1:

The patent introduces an intermediary processing stage between raw sensor data and neural network training. This intermediary layer involves preprocessing and feature extraction that filters out environmental noise and enhances meaningful fingerprint characteristics. By processing data through this intermediate representation, the model achieves both fast training and high accuracy by learning from cleaned, enhanced features rather than raw sensor data alone.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent implements feedback mechanisms during training that continuously monitor and adjust the learning process. By using feedback from recognition performance and loss functions, the model can correct errors and adapt to environmental variations, maintaining high accuracy even when trained on data collected under varying environmental conditions. The feedback loop enables the model to learn robust features that generalize well.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP4726674A1Fingerprint Anti-spoofing neural network training method and apparatus and fingerprint Anti-spoofing method and apparatus
Publication Date: 2026.04.15 SHENZHEN GOODIX TECH CO LTD
  • EP4726674A1 patent drawingFigure 1
  • EP4726674A1 patent drawingFigure 2
  • EP4726674A1 patent drawingFigure 3

AI summary

Embodiments of the present disclosure provide a method for training a fingerprint anti-counterfeiting neural network, a method for fingerprint anti-counterfeiting, an apparatus for training a fingerprint anti-counterfeiting neural network, and an apparatus for fingerprint anti-counterfeiting, comprising: obtaining a plurality of groups of training data, each group of the training data comprising: first raw domain data, second raw domain data, and third raw domain data; and training an initial classification network using the plurality of groups of training data to obtain a target classification network, wherein the initial classification network comprises a fusion subnetwork and a classification subnetwork, and for each group of the training data, the fusion subnetwork is configured to generate a first fingerprint matching pair based on a feature description matrix of the first raw domain data and the second raw domain data, and generate a second fingerprint matching pair based on a feature description matrix of the first raw domain data and the third raw domain data, and the classification subnetwork is configured to perform fingerprint classification and recognition based on the first fingerprint matching pair and the second fingerprint matching pair. The present disclosure solves the problem of low recognition accuracy rate of real and prosthetic fingerprints in related art, and achieves the effects of improving the recognition accuracy rate of real and prosthetic fingerprints.