Enhancing Biometrics with Auto Encoder: Accurate Finger Detection from Fingerprint Images

The manuscript introduced a novel method for the precise identification of individual fingers from fingerprint images. Our approach leverages a sophisticated deep learning framework centered on Auto Encoders. The primary objective of our method is to maximize the capabilities of Auto Encoders, allowing for the comprehensive analysis of fingerprint data, the discernment of latent patterns, and the extraction of meaningful features. Our research is founded on a rigorous evaluation using the SocoFing dataset, a publicly accessible repository containing a diverse array of fingerprint images sourced from various individuals and finger orientations. This dataset has been intentionally designed to encompass variations in image quality, noise levels, and rotational factors, thereby serving as a robust benchmark for assessing the resilience and performance of our proposed methodology. Remarkably, our Auto Encoder-based finger recognition technique has achieved an outstanding accuracy rate of 94%. This exceptional performance is attributed to the method’s capacity to intricately capture the nuanced patterns and unique distinguishing traits associated with each individual finger. Consequently, our approach excels in the accurate identification of fingers, even in the face of challenging and adverse conditions.