ARCHIVES
VOL. 12, ISSUE 1 (2026)
Adaptive fingerprint feature extraction using reinforcement learning-based filter selection
Authors
Dr. Sureshbabu N
Abstract
Fingerprint recognition systems depend on the
extraction of discriminative ridge features to achieve high accuracy under
diverse imaging conditions. Traditional feature extraction methods, which
utilize fixed convolutional or Gabor filters, often fail to adapt to
distortions, such as pressure variations, smudging, and partial prints. This
study introduces a novel Adaptive Fingerprint Feature Extraction (AFFE)
framework that employs a Reinforcement Learning-Based Filter Selection (RLFS)
mechanism to dynamically select the optimal filter for each fingerprint region.
The method integrates a Deep Q-Learning agent with a multi-scale convolutional
backbone to enhance ridge clarity, maintain orientation consistency, and
improve minutiae localization. Experiments conducted on the FVC2002 and FVC2004
datasets demonstrated that the proposed RLFS approach reduced the equal error
rate (EER) by 13.5% and increased recognition accuracy by 8.3% compared to
conventional CNN- and Gabor-based methods. The system exhibits strong
robustness against noise, partial prints, and distortions, making it suitable
for real-time biometric authentication.
Download
Pages:1-6
How to cite this article:
Dr. Sureshbabu N "Adaptive fingerprint feature extraction using reinforcement learning-based filter selection ". International Journal of Research in Advanced Engineering and Technology, Vol 12, Issue 1, 2026, Pages 1-6
Download Author Certificate
Please enter the email address corresponding to this article submission to download your certificate.