Abstract

Fingerprint recognition has emerged as one of the most reliable biometric authentication techniques due to its uniqueness permanence universality and resistance to forgery. In recent years growing digitalization and security requirements across banking e-governance telecommunications and mobile devices have accelerated the need for fast accurate and computationally efficient fingerprint recognition algorithms. Principal Component Analysis (PCA) traditionally used for dimensionality reduction in pattern recognition tasks has been extensively applied to fingerprint classification and identification due to its ability to capture dominant variance directions and convert high-dimensional fingerprint images into compact feature vectors. This research paper explores the integration of PCA with multiple distance measures—including Euclidean Manhattan Mahalanobis and Cosine metrics—to analyze how different similarity calculations influence recognition accuracy computational speed and robustness