Fingerprint-Based Blood Group Detection Using CNN and KNN: A Comparative Study

Abstract: For reasons such as discharge, legal processes, and emergencies, it is crucial to accurately determine blood group. Taking blood samples is at the heart of conventional wisdom, but it may not always be the best course of action. For the purpose of non-invasive blood grouping utilising fingerprint biometric methods, this research compares two algorithms, CNN and KNN, using a database of 500 fingerprint pictures labelled A, B, AB, and O (blood group). Achieving a remarkable 92.4% performance, CNN is able to record intricate spatial hierarchies and fingerprints using pattern layers in conjunction with ReLU activation. Performance has been enhanced by making minor adjustments to the photos. Because it relies on hand-crafted features and Euclidean distance, KNN—which achieves an accuracy of 76.8%— fails in a high-dimensional feature space. The error analysis shows that incompleteness and fingerprints are the key reasons CNN is poor. Due to its sensitivity to noise and overlaps, KNN demonstrates a higher level. A quick and effective non-invasive way to identify blood groups has emerged from research into the potential uses of convolutional neural networks (CNNs) in forensics, portable diagnostic devices, and automated blood transfusion management systems. The database will be expanded substantially and hybrid models will be used for improved performance in future studies.By conducting a thorough analysis of CNN and KNN, this work expands upon the idea of biometric blood group identification. The results show that CNN is more suited for this task than other methods because of its high noise power, which allows it to extract features more effectively. Our database will be further expanded in future studies, and hybrid models that combine the best features of several algorithm models to improve performance will be further investigated.

Keywords: Blood Group Detection, Fingerprints Biometrics, Machine Learning, CNN, KNN, Image Processing, Comparative Analysis