A new approach utilizing DTCWT for feature extraction and ANN for classification to improve fingerprint recognition

The authentication of biometric data using fingerprints is increasingly crucial in today's security landscape. Despite advancements in fingerprint identification algorithms, challenges remain in matching incomplete prints. Fingerprints, known for their uniqueness and long-term durability, are widely used in biometric research, including this study. While existing literature presents successful strategies, there is still room for integrating these methods into more advanced algorithms. As technology and security demands evolve, the need for efficient, cost-effective, and reliable systems has driven researchers to develop robust algorithms. This study demonstrates a direct correlation between test and database images, with a particular focus on forensic fingerprint evidence submitted in the court of law, which must be supported by scientifically validated methods. Dual-Tree Complex Wavelet Transforms (DTCWT) is effective in capturing intricate fingerprint details by decomposing images into multiple levels while preserving directional information, making it highly suitable for fingerprint feature extraction. When combined with Artificial Neural Networks (ANNs), which excel in learning complex patterns and classifying data based on feature vectors, The identification process utilizes a combination of DTCWT for feature extraction and ANNs for classification, constructing a feature vector that forms the basis of a multiclass classifier. Using the National Institute of Standards and Technology, Special Database-4 (NIST SD-4) for evaluation, the proposed approach surpasses notable machine learning algorithms, such as K-Nearest Neighbor, Decision Tree, Random Forest, and Support Vector Machine, based on performance metrics. This approach has proven particularly effective in handling incomplete or low-quality fingerprints, addressing a major challenge in biometric authentication systems, Utilizing DTCWT and ANN in fingerprint recognition is considered a significant advancement, offering a more scalable and accurate solution for security applications, including forensic analysis and biometric identification systems. Impact Statement. The surge in information technology and the demand for heightened security have compelled researchers and scholars to seek efficient, cost-effective, accessible, and reliable methods for individual identification. Conventional fingerprint matching algorithms struggle to preserve feature information, hindering their ability to handle partial input data and perform multiple computational functions simultaneously. The adoption of Man and Machine Intelligent systems emerges as a potential solution, warranting further in-depth research. This investigation centers on deploying ANN and gauging its effectiveness using assessment metrics like F_Score, Specificity, Precision, Accuracy, and Sensitivity. It aims to compare these results with the currently prevailing machine learning algorithms and classifiers.