Reinforcement Algorithm-Guided ROI Extraction of Fingerprint Biometric Data

Fingerprint recognition is a widely used biometric technique for personal identification. A crucial step in the fingerprint recognition process is feature extraction, where the unique characteristics of the fingerprint are identified and used for recognition. One of the challenges in fingerprint feature extraction is to select a region of interest (ROI) that can provide stable and reliable features for recognition. The existing techniques select an ROI around the core point for feature extraction. However, it is tough to detect the core point in noisy or low-quality images accurately. This paper proposes a novel deep learning-based method for fingerprint ROI selection that enables stable feature extraction. Furthermore, this method does not require cumbersome pre-processing steps such as fingerprint alignment or segmentation. The proposed fingerprint ROI selection method consists of two steps. In the first step, novel deep learning estimates the fingerprint region, including the core point. Next, reinforcement learning is used to optimize the region obtained from the previous step to get the required ROI. The approach’s effectiveness has been tested with standard fingerprint databases such as SPD2010, FVC2002 DB2, and synthetic fingerprints. Experimental results show that our proposed method effectively identifies a stable ROI for feature extraction. The proposed method can be applied to various fingerprint recognition applications.