Fingerprint Recognition With Early Block Termination Against Spoofing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Fingerprint recognition systems are vulnerable to attacks using counterfeit fingerprint images, and existing methods struggle with high latency in resource-constrained devices.
Innovation Solution
A fingerprint recognition method that divides the input fingerprint image into multiple blocks, calculates scores using a neural network, and terminates calculations based on preset conditions to reduce latency and improve real-time performance, utilizing a depthwise separable convolutional network with hardware acceleration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all image blocks are processed through the neural network to ensure recognition accuracy, then recognition accuracy is improved, but processing time increases
Solution Approach 1:
The patent applies partial action by processing only N image blocks (where N ≤ M) through the neural network instead of all M blocks. The system calculates scores for N blocks and uses early termination conditions to determine fingerprint legitimacy without evaluating every single block, thus reducing processing time while maintaining sufficient accuracy through selective sampling and iterative scoring.
Solution Approach 2:
The patent segments the fingerprint image into M discrete image blocks and processes them in an iterative manner. By dividing the image into smaller units and evaluating them sequentially with early termination capabilities, the system can achieve accurate recognition by processing only the necessary subset of blocks rather than the entire set, balancing accuracy and processing time.
2Reliability
If more image blocks are processed to improve recognition reliability, then reliability is improved, but device resource consumption increases
Solution Approach 1:
The system processes only N image blocks (N ≤ M) through the energy-intensive neural network instead of all M blocks. The early termination conditions allow the system to achieve reliable recognition by evaluating a sufficient subset of blocks, significantly reducing device resource consumption while maintaining recognition reliability through the iterative scoring mechanism.
Solution Approach 2:
The patent changes the parameter of processing quantity from fixed (all M blocks) to variable (N blocks where N ≤ M). By dynamically adjusting the number of processed blocks based on early termination conditions, the system optimizes the balance between recognition reliability and device resource consumption, processing only as many blocks as necessary to achieve confident recognition.
3Productivity
If the system processes fewer image blocks to reduce latency, then real-time performance is improved, but recognition accuracy may deteriorate
Solution Approach 1:
The system processes N image blocks (N ≤ M) which is fewer than the total M blocks, thereby improving real-time performance by reducing latency. The early termination conditions ensure that processing stops as soon as sufficient evidence is gathered, preventing unnecessary delays while maintaining recognition accuracy through the iterative scoring mechanism that accumulates evidence from processed blocks.
Solution Approach 2:
The patent implements feedback through the iterative scoring mechanism where each processed image block contributes to a cumulative sum score. The system continuously monitors the accumulated score against termination conditions, providing feedback that determines whether to continue or stop processing. This feedback loop ensures that accuracy is maintained by processing enough blocks to achieve confident recognition while improving real-time performance by stopping early when sufficient evidence is obtained.
Data Source
AI summary
A fingerprint recognition method includes receiving an input fingerprint image; dividing the input fingerprint image into M image blocks, wherein M is an integer greater than 1, calculating a score of N image blocks among the M image blocks by inputting successive ones of the M image blocks into a neural network for fingerprint recognition until the score of the N image blocks meets a preset condition, wherein the score is a sum score, and N is an integer less than or equal to M, and terminating calculating the scores of remaining image blocks among the M image blocks, and recognizing a legitimacy of a fingerprint in the input fingerprint image.