Comparative Study of Selected Feature Extraction Techniques for Fingerprint Recognition

In a fingerprint recognition system, feature extraction involves the actual process of transforming raw acquired fingerprint data images into a more useful representation for analysis or classification. The technique encompasses the identification and extraction of relevant characteristics from fingerprint image data. Feature extraction remains an effective method used to reduce the number of resources needed without losing vital information, and it has played a key role in improving the efficiency and accuracy of machine learning models or pattern recognition. This study conducted a comparative analysis of some selected feature extraction techniques for fingerprint recognition. Three feature extraction approaches, Discrete Wavelet Analysis (DWT), Minutiae, and Principal Component Analysis (PCA), were applied to extract features. The classification of extracted features was done using the Manhattan distance classifier. Evaluation of the developed fingerprint recognition system was achieved using two datasets: NIST DB4-Fingerprint Pattern and FVC2004 Dataset. The experimental results showed that the three selected feature extraction techniques and two fingerprint image datasets obtained results which revealed that the best accuracy value of 85.40%, was obtained in PCA, and the best FRR value of 9.80% was obtained in PCA, and best ERR value of 8.60% was recorded in PCA. Finally, it was shown that the performance of PCA was the best of the three feature extraction techniques for a fingerprint recognition system.