Fingerprint Image Rectification Using Neural Network Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Fingerprint sensor systems in devices often experience performance degradation due to defects such as scratches or cracks in the display stack or protective layers, leading to increased false rejection rates and reduced user experience.
Innovation Solution
A system that includes a fingerprint sensor, a control system, and a memory system to identify defective areas, determine if fingerprint image data can be rectified, and perform a fingerprint image reconstruction process using a neural network to mask and reconstruct images from defective areas, thereby improving sensor performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If fingerprint sensor systems are used with defective areas (scratches or cracks in display stack or protective layers), then the device can continue to operate, but false rejection rates increase and performance degrades
Solution Approach 1:
The system converts the harmful effect of defective areas into a beneficial outcome by detecting these defects and applying targeted masking and reconstruction algorithms. The neural network reconstructs fingerprint image data from defective areas by leveraging information from non-defective areas, thereby transforming the presence of defects into an opportunity to demonstrate the system's robustness and improve overall reliability through adaptive processing.
Solution Approach 2:
The system changes the processing parameters of fingerprint image data based on the detected defect locations. By dynamically adjusting the masking and reconstruction parameters according to the specific defect characteristics and positions, the system optimizes the rectification process to maintain authentication accuracy despite the presence of physical defects in the display stack or protective layers.
2Measurement precision
If fingerprint image data from defective areas is processed normally, then processing speed is maintained, but authentication accuracy decreases due to false rejections
Solution Approach 1:
The system performs preliminary defect detection and masking before the main fingerprint authentication processing. By identifying defective areas in advance and applying masks to exclude or reconstruct data from these regions, the system prevents corrupted data from degrading authentication accuracy. This preliminary action ensures that subsequent processing operates on cleaned or reconstructed data, maintaining both accuracy and efficiency.
Solution Approach 2:
The neural network acts as an intermediary between the raw fingerprint image data and the authentication process. It receives the original image data along with defect location information, processes this combined input through reconstruction algorithms, and outputs corrected fingerprint data. This intermediary processing step bridges the gap between defective input data and the requirements for accurate authentication without significantly impacting overall processing speed.
3Reliability
If the fingerprint sensor area is reduced to avoid defective areas, then authentication accuracy is improved, but the sensor area and user convenience are reduced
Solution Approach 1:
Instead of uniformly reducing the entire fingerprint sensor area, the system applies local quality adjustment by selectively masking or reconstructing only the specific defective areas while preserving the rest of the sensor area. This localized approach maintains the overall sensor area and user convenience while improving authentication accuracy by excluding or correcting only the problematic regions, thereby achieving high accuracy without sacrificing sensor size or user experience.
Data Source
AI summary
Some disclosed methods involve controlling a fingerprint sensor system to scan a portion of a user's digit and obtaining fingerprint image data corresponding to the portion of the user's digit. Some such methods involve obtaining defect data identifying one or more defective areas corresponding to an active fingerprint sensor area and determining whether the defect data indicates one or more defective areas corresponding to a touched portion of the active fingerprint sensor area from which the fingerprint image data were obtained. Some such methods involve determining, responsive to determining that the defect data indicates one or more defective areas corresponding to the touched portion, whether fingerprint image data corresponding to the one or more defective areas can be rectified and, responsive to determining that the fingerprint image data can be rectified, rectifying the fingerprint image data corresponding to the one or more defective areas.